~A G() 0~~~~~~"',. :2 3 . S Q0 0; o 0''t'e rlTˇ9E 'in''X P; mt& The Economics of Urbanization and Urban Policies in Developing Countries A World Bank Symposium The Economics of Urbanization and Urban Policies in Developing Countries edited by George S. Tolley Vinod Thomas The World Bank Washington, D.C., U.S.A. © 1987 by the International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, N.W., Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing April 1987 The findings, interpretations, and conclusions expressed in this study are entirely those of the authors and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. Maps that accompany the text have been prepared solely for the convenience of readers; the designations and presentation of material in them do not imply the expression of any opinion whatsoever on the part of the World Bank, its affiliates, or its Board or member countries concerning the legal status of any country, territory, city, or area, or of the authorities thereof, or concerning the delimitation of its boundaries or its national affiliation. Library of Congress Cataloging-in-Publication Data The Economics of urbanization and urban policies in developing countries. (A World Bank symposium) Conference held in 1983. Includes bibliographies. 1. Urbanization-Developing countries-Congresses. 2. Urban policy-Developing countries-Congresses. 3. Urban economics-Congresses. I. Tolley, George S., 1925- II. Thomas, Vinod, 1949- . III. Series. HT169.5.E25 1987 307.7'6'091724 87-2181 ISBN 0-8213-0786-X Foreword U rban problems in developing countries have be- take account of practical difficulties in implementing U come more acute in recent decades as people have projects and the realities of institutional settings. This flocked to cities, and the largest cities have been af- book places the best in economic analysis alongside fected the most. In coming years, as population growth real-world problems, contributes to fruitful interaction continues throughout the developing world, urban between the two, and provides analytical tools for im- problems promise to become increasingly severe. This proving policies. volume seeks to promote better understanding and eval- A unique contribution of this book is that the findings uation of policies designed to cope with these issues. as a whole provide a unified framework for earlier work To devise policies intelligently, we need to understand on urban problems. The studies build on progress to why urbanization has taken the forms we see today and date to make new contributions that should be of value what its future course is likely to be. Considerable prog- to policymakers, advisers, project analysts and other ress has been made recently in the economic analysis of practitioners, academicians, and observers of urban urbanization as a whole and its facets, including employ- problems. ment, housing, transport, pollution, and poverty. This The book consists of four parts, which are introduced book draws together studies of the causes of observed by an overview. Part I provides the economic underpin- urbanization patterns and builds on them to provide a nings for understanding urbanization. Comparative better foundation for policy analysis. advantage in urban and agricultural production, the Almost all countries have grappled with policies to relations between urban and rural poverty, and the control urbanization and manage growing cities more effects of total population growth are considered, and effectively. This book reviews the accumulating experi- the ways in which these and other factors have led to ence in both market and centrally planned economies observed differences in country urbanization experi- and provides a critique of successes and failures. ences are brought out. Urban policies are of two main types. The first at- Part 1I provides a framework for policy analysis. Cor- tempts to influence the degree of urbanization and the rection of market failures, redistribution of income, and distribution of population among regions and cities. In other goals are considered as bases for policy. Policy addition to surveying country experiences with such approaches in both market and planned economies are policies, this book provides methods for evaluating reviewed. them. The remainder of the book draws on the first two parts A second type of urban policy attempts to make the to derive tools for the evaluation of policies. Part III best of urbanization by coping with the urban manage- surveys national policies concerning urbanization and ment problems encountered during economic develop- gives examples of policies that encourage or discourage ment, particularly as they are aggravated by urban concentration of a country's population in one or a few poverty. Policies of the second type must be based on large cities. Experiences in Brazil, Colombia, the Repub- sound economic analysis of urban functions and must lic of Korea, and Venezuela are evaluated. Part IV ex- v vi Forewvord amines urban management problems, especially the Bank's accumulated knowledge of urban policy. It is provision of urban services and the maintenance of a based in part on a conference at the University of Chica- good or at least tolerable environment in large cities. go in 1983 in which experts from the Bank and from The emphasis is on techniques of quantitative analysis universities throughout the world participated. and on project management. Housing, transport, urban finance, and pollution control are considered. A con- Anthony A. Churchill cluding chapter draws policy lessons and makes sugges- Director tions. Water Supply and Urban Development Department The book, which contains several contributions that The World Bank originated from work at the World Bank, distills the June 1986 Contents Contributors xi Abbreviations xii 1. An Overview of Urban Growth: Problems, Policies, and Evaluation 1 George S. Tolley and Vinod Thomas Patterns of Urbanization 2 Urbanization and Economic Development 4 Sources of Future Urbanization 5 Economic Causes of Urban Problems 5 Urbanization Policy in Market and Mixed Economies 7 Urbanization Policy in a Centralized Economy 7 Concentration and Decentralization Policies 8 Addressing Urban Problems 10 Notes 12 Bibliography 12 PART 1. UNDERSTANDING URBANIZATION IN THE DEVELOPING WORLD 13 The Determinants of Urbanization 13 Urbanization in Representative Developing Economies 14 Conclusion 14 Bibliography 14 2. Urbanization and Economic Development 15 George S. Tolley Demand and Supply Factors 16 A More General Explanation 21 Conclusion 29 Bibliography 30 3. What Drives City Growth in the Developing World? 32 Allen C. Kelley and Jeffrey G. Williamson The Limits on City Growth 33 Predicting the Past 35 Understanding City Growth: Some Major Influences 38 The oPEc Watershed and Recent Growth Trends 41 Conclusion 43 Notes 43 Bibliography 45 vii viii Contents PART II. A FRAMEWORK FOR URBANIZATION POLICY 47 Urban Problems and Policy Implications 47 The Scope of Urbanization Policy 47 A Case of Central Planning 48 Bibliography 48 4. Market Failures as Bases of Urban Problems 49 George S. Tolley Urbanization and Poverty 49 Urbanization as a Source of Growth: An Illusion 51 Has Overurbanization Occurred? 52 Primal Cities 56 Economic Policy and Urbanization 57 Bibliography 58 5. Urban Development Policies in Developing Countries 60 Bertrand Renaud Basic Trends in Urban Policy 60 Urbanization Policies 63 Analytical Issues 65 The Direction of Urbanization Policies 66 Conclusion 70 Notes 71 Bibliography 71 6. Urban Policy in Centralized Economies: China 73 William L. Parish City Size and Stability 73 Useful and Productive Employment 76 Health Care and Housing 79 Conclusion 81 Notes 82 Bibliography 83 PART III. CONCENTRATION OR DECENTRALIZATION: EVALUATION OF POLICIES 85 Impetus to Urban Concentration 85 Locational Choice and Decentralization 85 Impact of Decentralization Policy 86 Bibliography 86 7. The Analysis of Urban Concentration and Decentralization: Thie Case of Brazil 87 J. Vemon Henderson Urban and Regional Patterns 87 Government Influence on Industrial Location 88 Sources and Magnitudes of Economies of Scale 90 Environmental Considerations in Decentralization 92 Conclusions and Policy Implications 93 Notes 93 Bibliography 93 8. Employment Location and Spatial Policies: Colombia and Korea 94 Kyu Sik Lee Observed Patterns of Employment Location 95 Behavioral Underpinnings of Location Choice in Manufacturing 98 A Framework for Evaluating Policy Effects 101 Policy Implications 102 Notes 103 Bibliography 103 Contents ix 9. Industrial Deconcentration Policy: Venezuela 106 Benjamin Reif Has Deconcentration of Industry Taken Place? 106 Is Deconcentration a Result of Policy? 109 Other Factors 110 Conclusions 118 Notes 119 Bibliography 119 PART IV. MANAGING THE CITY 121 Toward Greater Fiscal Efficiency 121 Urban Housing 121 Urban Transport 122 Urban Services and the Environment 122 Improving the Efficiency of Shelter Projects 123 Bibliography 123 10. Intergovernmental Fiscal Relations in Developing Countries 124 Roy Bahl and Johannes Linn Urban Government Finance as a National Problem 124 The Urban Fiscal Gap 125 Expenditure Pressures 125 Revenue Constraints and Opportunities 129 Policy Options for Urban Fiscal Reform 130 The Politics and Prospects of Urban Fiscal Reform 131 Notes 132 Bibliography 133 11. Housing Demand in the Developing-Country Metropolis 135 Gregory K. Ingram The Price Term in Housing Demand Equations 136 Housing Demand and Workplace-Based Price Variation 137 The Setting and the Data 138 The Hedonic Price Equations 139 The Housing Demand Equations 140 Aggregate Estimates of Income Elasticities 143 Conclusion 144 Notes 144 Bibliography 144 12. Urban Transport Policy: Colombia 145 Alvaro Pachon Operations 145 New versus Old Buses 148 Buses or Busetas? 150 Summary and Conclusions 153 Notes 154 Bibliography 154 13. Evaluating Pollution Control: The Case of Sao Paulo 156 Vinod Thomas Pollution: Problems and Policies 156 The Costs of Pollution Control 157 The Benefits of Pollution Control: Health Effects 159 Policy Implications 162 Notes 164 Bibliography 164 x Contents 14. Improving the Effectiveness of Urban Projects 166 Douglas H. Keare Project Efficiency and Effectiveness 167 Some Implications for Project Design and Policy Formulation 174 The Contributions of Evaluation 178 Notes 179 Bibliography 180 15. Urban Economic Policy: Directions for the Future 181 George S. Tolley and Vinod Thomas Urbanization Problems in Mixed Economies 182 Lessons under Central Planning 182 Urban Concentration and Urban Policy 183 Bibliography 184 Contributors Affiliations are those at the time of the conference. Roy Bahl Professor of economics and public administration and director of the Metropolitan Studies Program, Maxwell School, Syracuse University, New York J. Vernon Henderson Professor of economics, Brown University, Providence, Rhode Island Gregory K. Ingram Economic adviser, Water Supply and Urban Development De- partment, The World Bank Douglas H. Keare Chief, Education Division, Latin America and the Caribbean Regional Office, The World Bank Allen C. Kelley James B. Duke professor of economics, Duke University, Durham, North Carolina Kyu Sik Lee Senior economist, Water Supply and Urban Development De- partment, The World Bank Johannes Linn Senior economist, Country Programs Department, East Asia and Pacific Regional Office, The World Bank Alvaro Pach6n General Manager, Corporaci6n Centro Regional de Poblaci6n, Bogota, Colombia William L. Parish Professor of sociology, University of Chicago, Chicago, Illinois Benjamin Reif Professor, Faculty of Architecture and Urban Planning, Central University of Venezuela, Caracas Bertrand Renaud Senior economist, Water Supply and Urban Development De- partment, The World Bank Vinod Thomas Senior economist, Latin America and the Caribbean Regional Office, The World Bank George S. Tolley Professor of economics, University of Chicago, Chicago, Illinois Jeffrey G. Williamson Laird Bell professor of economics, Harvard University, Cam- bridge, Massachusetts Note: John Crihfield made a major contribution to the shaping of this volume. xi Abbreviations CEDE Centro de Estudios para Desarrollo Econ6mico (Colombia) CETESB Companhia Estadual de Technologia de Saneamento Ambiental (Brazil) CFT Corporaci6n Financiera del Transporte (Financial Transportation Cor- poration) (Colombia) cis Centro de Informacoes de Saude (Brazil) DANE Departamento Administrativo Nacional de Estadistica (National Statis- tics Department) (Colombia) EMPLASA Empresa Metropolitana de Planejamento de'Sao Paulo (Brazil) FEDESARROLLO Fundaci6n para la Educaci6n Superior y el Desarrollo (Colombia) FSDVM Fundaci6n Salvadorena de Desarrollo y Vivienda Minima (El Salvador) GDP Gross domestic product GNP Gross national product GSP Greater Sao Paulo Metropolitan Area IDRC International Development Research Centre (Canada) INTRA Instituto Nacional de Transporte (Colombia) isic International Standard Industrial Classification LCC Lusaka City Council (Zambia) MIT Massachusetts Institute of Technology OECD Organisation for Economic Co-operation and Development OPEC Organization of Petroleum Exporting Countries RDE Representative developing economy SEADE Fundac,o Sistema Estadual de Analises de Dados Estatisticos (Brazil) SEPLAN Secretaria de Economia e Planejamento (Brazil) TFPG Total factor productivity growth xii 1 An Overview of Urban Growth: Problems, Policies, and Evaluation George S. Tolley Vinod Thomas The world has been urbanizing rapidly for a long hibition of further industrial concentration-are in- T time and shows every sign of continuing to do so: tended to decentralize economic activity from large, more than 40 percent of the world's population today are congested urban agglomerations. The stated objectives urban dwellers. Dramatic urbanization over the past two include the need to reduce the economic and managerial or three decades has been concentrated in the develop- costs of overcrowding and improve regional and urban- ing countries, where the urban population has been rural income distributions. At the same time, other expanding at twice the rate observed in the countryside. economic policies-for example, credit subsidies, im- In fact, the world's biggest cities have increasingly been port protection, and urban food subsidies-more often in the developing nations, and population concentra- than not indirectly promote urban concentration. tions in such urban centers as Mexico City, Sao Paulo, Awareness of such effects has prompted some policy- and Shanghai have reached extraordinary levels. makers to reconsider these implicit spatial policies. The rapid growth of cities and their large sizes not- This volume is a response to the increasing preoc- withstanding, the urban share of total population in the cupation with urbanization in developing countries and developing countries as a whole is less than one-third, to the associated policy issues. In part, the book is compared with more than three-fourths in industrial concerned with an understanding of why and how ur- market economies. Furthermore, the current pace of banization occurs. An examination of the patterns of urbanization in developing countries is not un- urbanization among countries and over time puts in precedented: the urban population in some of today's perspective the more recent developing-country experi- advanced economies grew even faster in the past. The ences and brings out similarities in urbanization in difference lies in the much larger absolute sizes of a few different countries during economic development. Ex- cities and of urban agglomerations in the developing planations of the processes of urbanization are also countries today. The urban populations constitute a essential to better appreciate the benefits and costs con- relatively low share of the national totals, however, as a nected with those processes. These concerns are linked result of significant population increases from large to another goal of the book, an evaluation of urbaniza- bases in the rural areas. tion policies. Policy initiatives that attempt superficially Whether or not the current pattern of urbanization in to counter deep-rooted urbanization trends can be developing countries is alarming when viewed from a costly; policies to address the underlying causes of con- historical, spatial, and cross-country perspective, it cem and ameliorate urban (as well as rural) problems greatly concerns most governments in those countries. directly can be beneficial. Consequently, some countries follow policies intended Linn (1983) and Renaud (1981) have made important to slow down urban concentration and even reverse contributions in this area. Renaud compares macroeco- rural-to-urban migration. Some measures-such as the nomic, urbanization, and interurban policies in develop- creation of new cities and the discouragement or pro- ing countries with those in the developed countries and 1 2 George S. Tolley and Vinod Thomas discusses the need to correct the undesirable spatial Historical Comparisons effects of national economic policies, make internal management of cities more efficient, and increase eco- Some argue that developing-country urbanization to- nomic efficiency by eliminating barriers to mobility of day is qualitatively different from the historical pattern resources and dissemination of innovations. Linn fo- in today's developed countries. The main difference re- cuses in greater detail on intraurban issues and on how lates to the absolute levels of urbanization today, which to increase the efficiency and equity of growing cities. in good measure are based on large overall populations. He covers a wide range of areas-urban employment, High levels of absolute poverty are also associated with income distribution, transport, housing, and social ser- urbanization. During 1970-82 the developing countries vices-and evaluates the effectiveness of such policy experienced a more rapid urbanization than did the instruments as public investment, pricing, taxation, and industrial countries (table 1-1). The orders of magnitude regulation. are even more striking between 1950 and 1980, when This study focuses more sharply on the policy issues the urban areas of the developing countries (excluding connected with concentration and decentralization. China) absorbed nearly 600 million additional people- Evaluation methods and quantitative assessments of twice the total number of urban dwellers in industrial policy effects have been drawn from the experiences of countries at the beginning of that period.' different countries. The book complements the more If the recent trends continue, an additional 1 billion general urban review of Renaud and the city-specific urban dwellers may be added in developing countries by findings of Linn in that it seeks to sharpen the economic the end of this century. The magnitudes are historically analysis and evaluation of urbanization policies at the unique. Unprecedented concentration of people is likely national and city levels. A unique aspect of this book is to prevail in several cities in the developing countries its attempt to deepen our understanding of the eco- which are already among the biggest urban agglomera- nomic benefits and costs of urbanization and urban tions in the world. According to U.N. projections, Mex- policy interventions without denying the contribution of ico City and Sao Paulo each may contain more than 25 other disciplines to an understanding of urbanization. million people by 2000, closely followed by Shanghai, Beijing, and Rio de Janeiro; of the twenty-five cities that are likely to have more than 10 million people, twenty are expected to be in what are now considered develop- Patterns of Urbanization ing countries. The magnitude of this growth is likely to compound the problems of managing cities and of In comparisons of urbanization it must be kept in addressing absolute poverty and unemployment. mind that countries do not report their urban popula- Although the sheer magnitude of urbanization and of tions uniformly. The reasons for nonuniformity range associated problems in developing countries is phe- from use of different criteria in defining urban areas to nomenal, some other salient aspects of the present ur- differences in concepts and in the accuracy of statistics banization pattern do not depart markedly from past among countries. Furthermore, global data for the experience. A comparison of trends over many years urban sector usually gloss over important intraurban reveals no dramatic changes in the growth rate of the differences among countries, including variations in the world's urban population. The share of the world's definition and nature of large as against small cities. population in urban areas has increased steadily, from Statistics on urbanization should therefore be used with an estimated 3 percent in 1800 to more than 40 percent caution and perhaps only for broad comparisons. today, but the percentage change per decade in the Table 1-1. Urban Population: Share and Growth, by Country Group (percent) Urban population Average annual Average annual as share of growth in urban compound total population population growth in urban share, Country group 1960 1982 1960-70 1970-82 1960-82 Low-income 17 21 4.1 4.4 1.06 Middle-income 33 46 4.4 4.2 1.67 Industrial market 68 78 1.9 1.3 0.69 Industrial nonmarket 48 62 2.6 1.8 1.29 Source: World Bank (1984), annex table 22, p. 260; the last column is derived from the first four. An Overviewv of Urban Growth 3 proportion that is urban has remained around 16 per- largest city in 1980. In the middle-income countries the cent during this period.2As seen in table 1-1 the develop- proportion was substantially higher, 29 percent. In the ing countries today are adding to their urban popula- industrial market economies, however, it was only tions more rapidly than are the developed countries, but somewhat higher than in the low-income countries-18 these rates are not very unlike those observed during percent. Because of the high representation of rapidly earlier periods of urbanization in what are now de- growing countries in the middle-income group, the veloped economies. Hence, the long-term processes figures lend some support to the observation that one under way are not totally surprising. way in which rapidly developing countries of today differ from countries that underwent development earlier is in Comparisons among Countries the tendency for their urbanization to be more heavily concentrated in large cities. In fact, the overcrowding in Behind the total figures on urban population lies a one or a few cities, and the visible poverty connected variety of individual country experiences. Latin America with this phenomenon, are of prime concern; the overall is by far the most urbanized among the developing level and rate of urbanization in the country as a whole regions; some two-thirds of its population live in urban are less prominent issues. centers. In contrast, low-income Asia and Africa are predominantly rural, with average urbanization levels of Some Generalizations 25 percent. Intraregional differences are also significant: the urban share is about 83 percent in Argentina and 46 Some relations between urbanization and economic percent in Ecuador: it is 24 percent in India and 12 development can be further distinguished with the help percent in Bangladesh. of simple descriptive regressions. For sixty-six low- and Notwithstanding these differences, some compari- middle-income economies for which data are available, sons seem to apply broadly across countries. Descrip- regressions have been run using the following indepen- tively, a clear and well-known association is observed dent variables: (1) the per capita income rank of the between income level and the percentage of population economy, (2) the percentage rate of growth of per capita which is urban. As shown in table 1-1, for countrieswith income, (3) the percentage rate of growth of total low incomes, 21 percent of the total population was population, and (4) zero-one variables that represent the urban in 1982. In contrast, 46 percent of the population region in which the economy is located. (For example, was urban in middle-income countries, and 78 percent for an economy in Asia the Asia variable takes on a value was urban for the higher-income industrial market of one and the values of all other regional variables are economies. A broad positive association between income zero for that economy.) levels and degree of urbanization also seems to hold within regions. (There are, of course, many exceptions Degree of Urbanization. When the independent vari- to the simple relation between income level and the ables listed above are used in a first-regression equation percentage of population that is urban.) to explain urbanization, as measured by the percentage Furthermore, growth in urban population appears to of the population that was urban in 1980, a high degree have a tendency to decline with rising incomes. This of association is found. The multiple regression coef- tendency is reflected in differences in the rate of urban ficient, R2, is 0.769. The most significant variable in population growth among countries with different in- explaining the percentage of the population that is comes. During 1970-82 urban population grew at 4.4 urban is the income rank. This finding corroborates the percent annually in the low-income countries and at a positive relation already noted between income and the much higher rate if China and India are excluded. In the degree of urbanization. The growth of total population middle-income countries urban population growth was has a significant negative effect in explaining the per- 4.2 percent annually, whereas for the high-income in- centage of the population that is urban. This result may dustrial market economies the urban population growth be attributable to slow population growth in higher- rate was only 1.3 percent a year. income countries, which, in view of the positive income- These regularities in urbanization in developing urban relation, also tend to be the most highly urban- countries might be expected as part of the development ized. The growth of per capita income has a negative, process. Some aspects of present-day concentration may though less significant, coefficient in explaining the nevertheless be considered especially noteworthy. A urban share of population. The most rapid percentage tendency often noted is that large portions of the popula- gains in per capita income occur in countries other than tions of rapidly developing countries tend to be concen- those which already have high levels of income and trated in one or a few cities. For the low-income coun- urbanization. Thus, countries with the most rapidly tries, on average, 16 percent of the population was in the rising incomes usually have not yet reached the highest 4 George S. Tolley and Vinod Thomas 6 George S. Tolley and Vinod Thomas which are discussed in case study detail in later chap- that initially make up a small part of the economy, most ters-and points out that urbanization and its accompa- often in the urban sector, in early stages of development. nying phenomena are products of wider development The rise in marginal productivity of all labor, including processes. Care must be taken in the design of measures unskilled labor, may be slow at first and then may and policies concerning urbanization to avoid extreme accelerate as a greater proportion of the economy de- corrective actions that may hamper development. velops. As the marginal productivity of labor rises in the course of development, the incomes of those at the low Urban Poverty end of the income scale also rise. Furthermore, since development increases the returns to education and The most widely observed and acutely felt urban prob- other forms of human capital investment, many persons lem in developing countries is the large numbers of poor who would otherwise be unskilled transform themselves and unemployed people in the cities. The extent of into skilled laborers and thus directly raise their in- poverty in the economy as a whole depends on the comes. The number of unskilled employees falls, and degree of economic development. Urban poverty is their marginal productivity consequently rises. viewed as part of overall poverty; the rural poor move to Urbanization, which accompanies the strengthening urban areas, which tends to broadly equilibrate real of comparative advantages in urban sectors with devel- incomes across locations. Poverty is related in great opment, can be associated with considerable visible pov- measure to the size of the low-skilled population, which erty and other urban problems, particularly as unskilled is distributed among urban and rural areas according to workers become displaced from traditional rural activi- economic and other considerations that affect the entire ties. Sustained and continued development over a long population. time would be desirable to ameliorate adjustment prob- Growth of production leads to growth in demand in lems and increasingly absorb the previously unskilled urban areas for all factors of production, including un- laborers into the developing economy. The origins of skilled labor. The change in the proportion of unskilled urban problems lie in inadequate and unsustained de- workers in urban areas depends on growth in demand velopment and rural-urban adjustment, but those are by for urban and nonurban commodities, on substitutabili- no means the only sources of difficulties. ties between unskilled workers and other factors of pro- duction, and on the comparative decline of so-called traditional production in urban and in rural areas. Tra- ditional production may be defined as production by Increasing concentration of economic activities and firms or households which, although they can modify people has been viewed as a reflection both of develop- their output and techniques in response to changes in ment and of economic deterioration. As already noted, prices of products and factors, are not viable in the face whereas the generally higher incomes in urban areas are of competition from modern domestic or foreign manu- associated with the benefits of urbanization, urban facturing or from modern farming enterprises. Tradi- poverty and unemployment and a host of problems tional production usually uses unskilled labor, and its associated with pollution and congestion are the most decline may release unskilled labor to other parts of the noted indicators of urban failure. Embedded in the ur- economy. Although some analysts explain economic de- banization process are elements that represent develop- velopment exclusively as a decline in traditional produc- ment and that deserve to be promoted, but urbanization tion, development is here viewed as a broader process also brings with it extemalities which can bring about which involves an accumulation and transfer of knowl- overcrowding. edge that could and probably would occur even if no Three types of externalities may be associated with units became unviable. The large-scale decline of indus- overurbanization. First, environmental externalities, tries that engage in traditional production may be an such as those connected with pollution and congestion, important accompaniment of development in some can mean that cities' sizes are larger than would maxi- countries-since the decline is accompanied by sub- mize national (or regional) income and welfare. Second, stitution of capital for labor and by introduction of new protected employment that maintains urban wages techniques-but in other countries this process may be above market-clearing levels may make cities larger or an unimportant detail. Other things being equal, if the smaller than they otherwise would be, depending on the decline in traditional production is greater in rural than elasticity of demand for labor. Third, the attractiveness in urban areas, the proportion of unskilled workers and of urban areas (because of the availability of free or hence of poor city residents will increase. subsidized public services and the advantages of proxim- Even in the absence of traditional production, de- ity to government activities) can lead to excessive urban- velopment may be concentrated in industries or firms ization. Going beyond these urban externalities, a An Overview of Urban Growth 7 hypothesis offered here is that the phenomenon of great of regulations in India that favored rural over urban urban concentration in one or a few cities is connected development has been the encouragement of the growth with the difficulty and relatively high cost of providing of the urban underground economy. The use of water intercity transport in developing countries, and that resources for regional development purposes in Malaysia more general infrastructure decisions also play a role. has been dramatic, and policy attention to leakages and regional multipliers has been greater than usual. The Republic of Korea, because of its shortage of land, has an Urbanization Policy in Market intense interest in land-use guidelines (see chapters 5 and Mixed Economies and 8), but in spite of prohibitions and incentives, ex- port-oriented growth policies and industrialization en- Some of the differences among today's developing sured that the largest urban centers, especially Seoul countries that merit attention are emphasized by Re- and Pusan, would continue to grow. Venezuela and naud in chapter 5. In many of today's developing coun- Brazil have pursued industrial deconcentration policies, tries additions to population are larger, income levels with limited success (chapters 9 and 13). In addition to are lower, and opportunities to relieve population pres- policy incentives or disincentives, the balance of labor sures through migration are more limited than in market considerations, economies of scale within an others. Urbanization in today's advanced economies pre- industry, and the effect of the total size of a city deter- sents a different set of issues. In some countries there is mine location decisions (chapter 7). a process of a slowdown and an eventual end to urbaniza- The appropriate internal management of cities is im- tion and the emergence of yet another industrial revolu- portant to the success of national spatial policies. Sub- tion based on new technologies which are less tied to Saharan Africa exhibits acute problems of city man- concentrated manufacturing centers. agement in the face of rapid urbanization, albeit from Chapter S provides a review of urban-related problems modest bases. Centralization of decisionmaking in Ni- in developing countries-regional inequalities, conges- geria, for example, has made it difficult to deal with local tion, pollution, and inadequate provision of urban ser- problems. In very large cities, policies to limit or stop vices-and a threefold classification of related urbaniza- population growth are not good substitutes for policies tion policies: national economic policies, such as import that address the urban bias and that directly seek to controls, that have spatial effects; explicit regional de- correct congestion and pollution and provide adequate velopment policies, such as investments in infrastruc- services. If other cities are not efficiently and effectively ture; and policies concerned with the management of managed, their chances of attracting industries and mi- cities. Problems and policy emphasis vary among coun- grants from the largest urban centers will be small. tries: they are influenced by the age-old structure of Regional policies can strengthen promising secondary cities in Asia, the rapid urbanization which has been urban centers through well-chosen, cost-efficient ac- going on for some time in Latin America, and the new tions, better policies for transport investment and urbanization in several parts of Africa, which started management, industrial estates policies, and more im- from a low base and is proceeding at unprecedented portant, the systematic development of organized infor- rates. mational networks, such as banking networks, indus- National economic policy sometimes produces sig- trial association networks, and better administrative nificant and unintended effects on urbanization which structures, between the secondary cities and the capital may outweigh any direct and intended effects of urban- region. Good city management and selected regional ization policies. The unintended spatial biases of nation- investments, combined with national economic policies al economic policies usually favor some urban centers that do not discriminate against rural areas, would do and are commonly generated through trade policies that much to alleviate the problems of urbanization. protect the manufacturing sector. For example, credit allocation, public investment, and pricing policies may Urbanization Policy in a Centralized Economy give preferential treatment to economic activities that are concentrated in a few cities and regions. The man- Parish (chapter 6) describes how China's socialist- agement practices of the central government and its leaders, since their coming to power in 1949, have tried regulation of economic activities, which make location to shape the nation's cities to avoid many of the urban close to the capital necessary or advantageous for firms, problems encountered elsewhere in the developing also contribute to the urban vortex. world. Policymakers planned to restrain the rate of ur- Explicit regional development policies sometimes banization, since rapid urban growth has often been attempt to favor decentralization, but they are not al- accompanied by an insufficient number of suitable jobs ways successful or efficient. An unintended by-product for new migrants from the countryside. They sought to 8 George S. Tolley and Vinod Thomas avoid the pattern found in other countries of rapid frustrated because the same level of education had growth of a few large coastal cities at the expense of guaranteed good jobs just a decade before, when educa- smaller interior cities. In addition, they tried to narrow tion was less common. This frustration-shared with the significant gaps in living standards between Chinese youths in many developing countries but perhaps felt cities and rural areas by creating secure jobs and guaran- more acutely in China because of the socialist promise of teeing such basic services as health, education, housing, secure jobs and rapid development-contributed to the and essential food supplies for both urban and rural outbreak of petty crime in the 1970s and continues to populations. fuel the social alienation of some youths. China's history since 1949 brings out both the poten- China's experience with the virtual elimination of tial and the disadvantages of urban development in a small informal service activities illustrates the necessity centrally planned economy. With greater control over for these types of activities in developing cities. Some economic resources than in the average market society, argue that because of high population growth rates and the government was able to shift investment funds to the difficulty of centralized provision of essential urban promote the development of medium over large and services, the informal service sector in poor societies is interior over coastal cities, which helped reduce regional essential and should be embraced rather than shunned. inequalities. Control over jobs and rationed consumer China's leaders seem to be moving closer to that view, supplies meant that for a time the government was able which entails a greater reliance on market forces, even to limit severely the growth of all cities, and funds that though they reject its extreme version. Smaller, more might have been spent on an elaborate urban infrastruc- makeshift work arrangements, organized ad hoc by ture for new migrants were spent instead on rapid in- neighborhoods and individuals, with lower rates of pay dustrial growth. and security, are now approved as a way of providing The government has been able to put the urban both employment and essential urban services. Further population to work. Most able-bodied women, and more efforts to improve the relative position of rural activities than half the urban population, are employed. Few of the are being made, and farmers are beginning to narrow jobs are part-time or likely to be lost tomorrow; they are the gap between average rural and urban incomes. primarily full-time jobs that promise to last for a full career. In these ways Chinese cities have avoided some of the severe problems of unemployment and employment Concentration and Decentralization Policies instability that have afflicted other developing-country cities. At different times and in different countries the Policy mistakes have, however, been made and have growth of large cities has been welcomed and deplored. led to some of the same problems found in other soci- Governments have used policy instruments to encour- eties. For instance, it was an error to downplay the role age location in major cities or to foster-sometimes of light industry and consumer services. Some develop- even to force-diffusion. Chapters 7, 8, and 9 examine ing-country cities may have too many people in informal the reasons behind spatial location policies in four coun- service activities, but China illustrates the problems of tries and the outcomes of those policies. the opposite extreme. The restricted opportunity for growth of light industry and the bureaucratic restraints En on small individual enterprises have fed problems of couragement of Concentration: Brazil youth unemployment. Although elimination of urban- The relative emphasis on concentration and on decen- rural income gaps was an objective, on balance a pro- tralization in Brazil has changed with the varying for- urban and proindustrial bias might nevertheless have tunes of the economy; concern about concentration has existed, as in many other countries. often diminished during times of economic difficulty. The unemployment problem was heightened by the Neither concentration nor decentralization may be un- rapid expansion of employment for women. Jobs were desirable in itself; government policies, however, may created during the 1960s and 1970s, but since those jobs have intended and unintended effects on industrial loca- were taken almost as frequently by women as by men, tion which affect economic welfare. In chapter 7 Hen- the need to create additional openings was greater than derson examines how government policies have in- in other developing societies. Furthermore, the rapid fluenced concentration in Brazil, and he attempts to expansion of secondary school education, as in many evaluate the desirability of such influences when the other developing societies, contributed to growing num- existence (or absence) of certain types of economies of bers of unemployed educated persons. China avoided scale in urbanization and industrialization is taken into this problem at the university level but not at the sec- account. The data base for the core of the paper relates to ondary level. The unemployed youths were particularly 1970, and much of the discussion therefore concerns the An Overview of Urban Growth 9 historical evolution of policies which essentially pro- with particular attributes. Hence it is important for moted concentration. Some aspects of the distinct ef- policymakers to understand the requirements of firms forts in the 1970s to achieve decentralization are cov- for attaining equilibrium at new locations and to be able ered in chapter 13. to assess the level and costs of government subsidies and Localization economies are found to be strong in infrastructure investment needed to meet such require- Brazil, and therefore agglomeration of firms into spe- ments. cialized cities-to take advantage of such benefits as The Bogota study did not test the effectiveness of efficiencies in labor markets and in services specific to explicit policy instruments, partly because such instru- an industry and greater specialization among firms ments were not implemented in that city. Nevertheless, within an industry-is advantageous. The present re- the behavioral underpinnings established in the study sults do not show any significant urbanization econo- provide clues as to which policy instruments are most mies at the scale of activities prevalent in the urban appropriate for influencing both the location choices of centers of the South and Southeast of Brazil in 1970. particular types of firms and the aggregate locational The rationale for efforts to encourage industrialization patterns. It is apparent that government policies in- of the largest urban areas rests on the putative net tended to influence employment locational pattems can benefits for heavy industries from locating in areas with be effective if they influence the site attributes which are a large general scale of economic activity. Henderson's important to firms. findings do not, however, support this rationale. Rather, Lee also examines urban policy in Korea. During the they indicate that efforts to limit or counter decentral- past decade various spatial policies to control the growth ization initiatives may not be desirable. of Seoul and to disperse its population have been im- In addition, negative extemalities in the form of en- plemented. For example, in 1971 the greenbelt sur- vironmental degradation could constitute grounds for rounding Seoul was established. Subsequently the 1977 actually promoting some degree of decentralization. But Industrial Location Act in effect prevented new manu- the size distribution of Brazilian cities is by no means facturing firms from locating within Seoul and enabled excessively skewed, and efforts to bring about decentral- the government to issue relocation orders to establish- ization or a different distribution of city sizes for its own ments already set up there. In 1977 the government sake may not be warranted. Nevertheless, the provision initiated a ten-year comprehensive plan for redistribut- of more uniform incentives to middle-size cities-which ing population and industry away from Seoul. could simply mean the elimination of any special incen- Several other developing countries have tried to de- tives, direct or indirect, for larger cities in the southern centralize economic activity away from the central city. region-coupled with environmental restrictions in the The economic desirability of decentralization policies highly damaged and built-up areas, could lead to an has not been established, and not much is known of their economically beneficial decentralization of activities. effects or their welfare implications. The key policy question is how to guard against spatial policies that are Industrial Mobility and. Decentralization:excessive in relation to prevalent decentralization Cndstr.iaa Moian trends, since such measures might result in serious welfare losses. In developing countries the lack of empir- Lee (chapter 8) documents the changing patterns of ical information on decentralization and policy effects employment location in Bogota and Cali, the first and does not yet permit the formulation of more efficient third largest cities in Colombia, and summarizes some spatial policies, but policies to decentralize population econometric work on location choices by manufacturing and economic activity are probably not good substitutes firms. The study also outlines a framework for measur- for better intemal management of city growth. For ex- ing policy effects and draws policy conclusions, particu- ample, the effect on air pollution or on traffic congestion larly in the context of Korean experience with spatial of reducing the population or employment in a large city policy. by a certain amount is likely to be small. The main phenomenon observed is the policymakers' frequent attempts to relocate industries from the tradi- A Decentralization Program: tional industrial districts of large cities to outer areas or Venezuela to smaller cities. The government's plans may include developing new industrial towns or estates or expanding Reif (chapter 9) uses the case of Venezuela to provide a existing ones to induce new firms or firms that are full-fledged evaluation of decentralization policy. The moving to settle in a desired area. In all cases imple- goals of Venezuelan policy have been to prohibit the mentation of such plans and programs requires the location of new manufacturing in Caracas and its sur- selection of particular types of industries to occupy sites roundings, to induce hazardous industries to relocate 10 George S. Tolley and Vmod Thomas and to encourage others to move to designated develop- reassign responsibility for urban services from local to ment areas, and to attract new manufacturing plants to central authorities, but Bahl and Linn (chapter 10) do the designated areas. Policy instruments include direct not recommend this approach. An alternative solution is financial and fiscal incentives, indirect benefits, and to increase local tax authority and make more use of negative incentives such as locational control. Reif finds property taxes and motor vehicle taxes. A third solution, that although little overall deconcentration took place also favored, is to encourage reliance on user charges from 1971 to 1978, new firms did tend to leave the (which have the advantage of being directly linked to country's dominant industrial area. services provided). Equity considerations and social It remains to be established, however, that the decon- costs (for instance, congestion) may, however, keep user centration was the result of government policy. An charges low. A fourth solution, to increase fiscal trans- analysis of the effects of financial, fiscal, and negative fers from central to local authorities, has many pros and incentives indicates that these had some, though not cons but might be recommended if proper encourage- major, effects on deconcentration. Reif investigates ment to cities can be combined with preservation of other factors and hypothesizes, among other things, local autonomy. Reforms in urban financial arrange- that firms which receive government support benefit ments in developing countries are proceeding slowly; from locations near government centers. A series of logit political factors contribute to the inertia. regressions indicates that wages, access to markets, a well-trained work force, union activity, and availability Housing Policy of water significantly affect firm location decisions, in contrast to the relatively less significant performance of In developing countries, housing is a major consump- government financial incentives. The work shows that tion category; it constitutes, on average, 15-25 percent strong economic tendencies must be overcome if any of total urban household expenditures. Housing is also appreciable decentralization is to be achieved. one of the most problematic areas in urban develop- ment. The most glaring aspect of the problem concerns sprawling squatter areas, delapidated shelter, and appalling lack of basic public services. Other dimensions Addressing Urban Problems of the problem are shortages of land, infrastructure services, off-site services, amenities, and employment Part IV of this book contains discussions of some of opportunities and the inadequate supply and rising cost the significant problems that accompany urbanization. of housing. Among the most pressing are management of urban For the overwhelming majority, housing is privately fiscal resources, housing, transport, and environmental provided. Public authorities, however, implement protection. licensing, building codes, zoning, and recordkeeping to ensure clarity in property rights. Governments almost Management of Urban Finance universally have some active interest in housing finance. Overt programs of public housing for low-income Growing fiscal problems in cities of the developing groups are common. Housing policy must account for countries have been caused in part by unprecedented people's desires and should maximize incentives for in- urban growth, for which many countries are ill- dividuals to expand the housing supply. Public interven- prepared, and by the growing demands on local services. tions should be limited to areas where the public sector The demand for local services is sensitive to increasing is best suited to perform, such as direct investment, population, especially of the poor. Positive income elas- pricing policy, and regulation. ticities for public services and demonstration effects To provide housing requires expertise in many areas. from the developed world also influence the demand for Ingram (chapter 11) stresses the need for understanding local services. Factors that affect the cost of local ser- the underlying demand for housing. He analyzes hous- vices include wages, labor unions, rising land and ing demand by renters and owners in Bogota and Cali, energy prices, inflation, and the increasing costs of Colombia, for 1972 and 1978. Hedonic price coefficients borrowing funds for large urban infrastructure invest- for housing attributes are used to estimate the cost of a ments. standardized unit of housing for each workplace, which The major revenue sources-taxes, user charges, and is then used as the price variable to estimate a demand external funding-tend to rise less than expenditures function for housing that includes other variables as during urbanization, and the result is fiscal and service well. The effect of income on the demand for housing is deficits. One solution to fiscal problems would be to highly significant, and its elasticity is in the upper end of An Overview of Urban Growth 11 the range 0.2-0.8. Price performs less well: its elasticity investments in buses, fare structures that favor new appears to be less than one. Other variables include age vehicles, and licensing requirements that restrict vehi- and sex of the household head, family size, and distance cle stocks are features of the Colombian system. to work. The dependence on public transport of a large part of the urban population, particularly the poor, heavily in- fluences fares and services and increases govemment involvement. The Colombian analysis indicates the Transport influences the rate and pattern of urban scope and problems of providing efficient services within development and presents multidimensional issues. In- this framework. frastructure in the form of roads and streets is publicly provided but entails private use. Urban transport plan- ning requires road layout and in some cases provision Urban Environment for commuter rail transit. These large-scale projects are Environmental problems are intimately related to examples of investments that need rigorous benefit-cost urbanization because growing pollution in the develop- analysis. ing countries has been caused largely by growth of The planning of transport and other large urban in- activity in urban areas. Concem has been heightened as frastructure investments is complicated not only by consciousness about pollution has spread worldwide. income distribution considerations but also by the dif- Thomas (chapter 13) analyzes policies for dealing with ficulty of taking account of the feedback effects of invest- the key tradeoff between reducing environmental dam- ments on urban development. If provision of services age, on the one hand, and, on the other, paying the cost merely follows demand, infrastructure may be a bot- of pollution control and maintaining industrial compet- tleneck to development, and development opportunities itiveness and growth. may be lost. But if infrastructure leads development and, In Brazil the capital and operating costs of air pollu- among other things, fosters new spatial patterns, several tion control equipment and spare parts can be used to problems arise. Infrastructure services that do not show how costs of pollution abatement vary among closely follow existing pattems carry the risk that de- producers. Where effluents per unit of output decline mand will not materialize as had been projected. As a with the size of a firm, it would be advantageous to minimum, benefit-cost analysis of urban projects might require less than proportionate abatement for larger introduce spatial modeling to show probable future res- producers. An offsetting effect, however, is economies of idential and business locations and to estimate urban scale in pollution control. The variation in pollution development with and without the proposed project. control costs by type and size of firm should be consid- Another aspect of the urban transport problem is the ered in devising policies that will achieve a given amount provision of public transport, which again involves in- of environmental control at the lowest possible cost. teractions between public and private decisions. In his The damages to health from the high degree of air analysis of public transport Pach6n (chapter 12) points pollution in Sao Paulo are large by any standards and in out that private organizations are responsible for most comparison with the relatively lower risks in outlying mass transit in Colombia, although local governments areas and in Rio de Janeiro. A regression of the mortality grant licenses and route authorizations. Disadvantages rate on the pollution level, population density, per capita can arise from the political allocation of routes; exam- income, hospital beds per person, and percentage of ples are parallel routes, duplication of service, and in- people sixty-five years of age and older yields a positive formation problems created by the multiplicity of and significant coefficient for particulates. For example, routes. The system is, however, flexible and appears to an annual increase of one ton of particulates per square be capable of handling the information problem. kilometer in the Greater Sao Paulo metropolitan area is Pach6n gives an economic rationale for preferring associated with an increase in mortality of twelve per- small, less capital-intensive vehicles (such as school sons per million. The analysis emphasizes that benefits buses, minibuses, and collective taxis) over large metro- from a given amount of pollution reduction in an area politan buses. The growth in the number of small buses will depend on the size of the population. in Bogota is a result of lower costs, shorter waiting The most advantageous tradeoff between reducing times, higher trip frequencies, and a greater income environmental damage and maintaining growth can be elasticity of demand for small-bus service. The desirabil- achieved by a policy that allows industrial and spatial ity of old and of new buses is also relevant; an analysis of variation in pollution control rather than mandating the age structure, operating costs, and profitability of uniform controls. A promising approach is to use emis- Colombian buses is provided. Subsidies that encourage sion taxes or pollution abatement standards that can be 12 George S. Tolley and Vinod Thomas adjusted to conditions in different industries and dis- 2. These estimates are based on Davis (1972), vol. 2, p. 51, tricts. and on more recent data from the World Bank. Evaluating Urban Projects Bibliography The findings of this book point to net benefits from selected policy interventions that directly address Andors, Stephen. 1978. "Urbanization and Urban Government selected urban problems. In this connection careful in China's Development: Towards a Political Economy of evaluations of intended and actual outcomes of urban Urban Community." Economic Development and Cultural programs would be helpful. Part of the work would be Change, vol. 26 (B), pp. 52545. better evaluations of urban projects, many of which Berry, Brian J. L., Peter Goheen, and Harold Goldstein. 1971. invole shlter Keae (capte 14)show tha rigrous "Metropolitan Area Definition: A Reevaluation of Concept involve shelter. Keare (chapter 14) shows that rigorous and Statistical Practice." In L. S. Bourne, ed., Internal evaluation of projects can assist in improving the effi- Structure of the City: Readings on Space and Environment. ciency and effectiveness of future endeavors in urban New York: Oxford University Press. policymaking and in project formulation and implemen- Davis, Kingsley. 1965. "The Urbanization of the Human tation. Although much of the paper deals with specific Population." Scientific American, vol. 213 (September), shelter projects financed by the World Bank, the lessons pp. 28, 40. drawn are likely to apply to policymaking as a whole. . 1972. World Urbanization 1950-1970. 2 vols. Popula- The projects are evaluated on the basis of eight tion Monograph Series 4 and 9. Berkeley, Calif.: University criteria: project design, the selection of project ben- of California Press. eficiaries, construction methods, materials loan pro- Hauser, Philip M., and Robert W. Gardner. 1982. "Urban Fu- grams, housing completion, occupancy, maintenance of ture: Trends and Prospects." In Philip M. Hauser and others, housing infrastructure, and community participation. eds., Population and the Urban Future. Albany, N.Y.: State Given a target group of beneficiaries and a policy objec- University of New York. tive (in this case, shelter), desirable projects are those Linn, Johannes F. 1983. Cities in the Developing World: Poli- that strive for efficient resource use through decentral- cies for Their Equitable and Efficient Growth. New York: ized decisionmaking. Keare stresses market solutions Oxford University Press. whenever possible and presumes that project partici- Mills, Edwin S. 1972. "Welfare Aspects of National Policy pants are the best judges of their own self-interests. In toward City Sizes," Urban Studies, vol.9, no. 1, pp. 117-24. general, projects should provide participants with suit- Mohan, Rakesh. 1979. Urban Economics and Planning Mod- able locations, secure tenure, and adequate credit but els: Assessing the Potential for Cities in Developing Coun- beyond these should leave most decisions to the partici- tries. Baltimore, Md.: Johns Hopkins University Press. pants. Advantages and disadvantages are associated with Renaud, Bertrand. 1981. National Urbanization Policy in De- construction projects, self-help requirements, housing veloping Countries. New York: Oxford University Press. standards, rentals, and restricted credit policies. The Richardson, Harry W. 1977. City Size and National Spatial costs of delayed occupancy and inadequate maintenance Strategies in Developing Countries. World Bank Staff Work- and the importance of project cost recovery are stressed ing Paper 252. Washington, D.C. in this context. United Nations. Population Division. 1977.Manual8. Methods for Projections of Urban and Rural Population. Population Studies 55. New York. Wheaton, W. C., and H. Shishido. 1979. "Urban Concentra- Notes tion, Agglomeration Economies and the Level of Economic Development." Working Paper. Cambridge, Mass.: MIT. 1. Much of the information in this section is from World World Bank. 1984. World Development Report 1984. New Bank (1984). York: Oxford University Press. Part I Understanding Urbanization in the Developing World Part I of this volume contains contributions by Tolley tion that occurs during development is intensified. A key and by Kelley and Williamson that provide a factor behind this phenomenon may be that the transfer framework for understanding urbanization processes. In of production techniques between nations appears to be chapter 2 Tolley views urbanization as a result of the less costly for urban than for agricultural products. In interplay between supply of and demand for urban goods addition to these factors which affect the proportion of a and services. Kelley and Williamson in chapter 3 adopt a country's population which is urban, population growth more elaborate general equilibrium model to analyze in the country as a whole increases the number of people city growth in the developing world. The two studies living in urban areas. focus on fundamental economic reasons for urbaniza- Following the above line of reasoning, a simple model tion and offer complementary insights into the process. of urban population growth would be based, among other things, on (1) total population growth, (2) urban productivity derived from nonlabor sources, and (3) rural productivity derived from nonlabor sources. An The Determinants of Urbanization application to sixty-six developing countries reveals the possible effect of those variables. As expected, total Chapter 2 begins with a simple model that introduces population growth contributes to the rate of urbaniza- the principal determinants of urbanization and consid- tion, but it accounts for only a 2 percent annual rate, and ers increasingly complex and more realistic situations, actual urbanization growth rates range up to 7 percent. Income elasticities of demand are higher for commod- Low rates of growth of nonlabor rural income are found ities produced in urban areas than for those produced in to be associated with high rates of urbanization; the rural areas, and this factor encourages urbanization. reason appears to be sluggish growth in agricultural Changes in income elasticities are hypothesized to occur productivity rather than land scarcity. More vigorous as an economy develops and to make for first a rising and increases in nonlabor inputs in urban areas provide then a falling rate of urbanization. Capital accumulation further impetus to urbanization. Relatively stronger and other reasons for changes in the use of labor affect productivity growth in urban nonlabor income sources, the number of workers needed to fulfill the demand for which arises from the relatively easier transfer of tech- urban and nonurban commodities and thus further niques among nations in urban as against rural produc- affect urbanization. tion, fosters urbanization at a rate faster than overall The progression from a closed-economy model to a population growth. Finally, in countries in which the more general open-economy model of urbanization level of urbanization is low-mostly the poorer na- based on comparative advantage underscores the im- tions-a given absolute addition to the urban population portance of a neglected reason for urbanization, spe- translates into a higher urban growth rate. The exis- cialization connected with international and domestic tence of many variations in the general patterns indi- trade. In countries that achieve development by spe- cates, however, that conditions in individual countries cializing in manufactured commodities, the urbaniza- play a role in urbanization. 13 14 Part I Urbanization in Representative tries as strong influences on urbanization in the coming Developing Economies decades. A general equilibrium model is used in chapter 3 to analyze city growth in developing economies. Prices of Conclusion outputs and inputs are flexible, and capital and labor are essentially mobile. The eight sectors of the model are The studies in part I draw attention to the importance manufacturing, modern services, informal urban ser- of the macroeconomic development of an economy, vices, low-quality urban housing services, high-quality trade policy, and relative prices and productivities in urban housing services, agriculture, informal rural ser- urban and rural areas as major influences on urbaniza- vices, and rural housing services. The nontradable ser- tion. Although these factors are not always considered in vices are important in determining spatial cost-of-living urban policy analysis, they may be highly significant. differences that affect migration behavior. Macroeconomic policy biases, particularly those that Savings by businesses (from after-tax profits), govern- affect trade, industry, and agriculture directly or in- ment, and households are allocated endogenously to directly, may also have significant impacts on urbaniza- investment in physical capital, human capital, and hous- tion. Some of these relations are further examined in ing. Exogenous variables are foreign-supplied capital, part II. the unskilled labor force, sectoral rates of total factor productivity change, prices of imported raw materials and fuels, and the terms of trade between primary ex- Bibliography portables and manufactured importables as influenced by commercial and trade policies. Beier, George, Anthony Churchill, Michael Cohen, and Ber- The results lead the authors to suggest that rapid rates trand M. Renaud. 1975. The Task Ahead for the Cities of of population growth are not the most important factor Developing Countries. World Bank Staff Working Paper 209. in rapid urban growth in developing economies. (This Washington, D.C. Also in World Development, vol. 4, no. 5 finding is not inconsistent with Tolley's view that overall (May 1976), pp. 363-409. population growth as such contributes to both urban Keesing, Donald B. 1979. Trade Policy for Developing Coun- popultiongrowh assuchcontrbute to oth rban tries. World Bank Staff Working Paper 353. Washington, and rural expansion.) Capital transfers to developing D.C. countries do not seem to have been significant. Nor is Kelley, Allen C., and Jeffrey Williamson. 1984. What Drives rural land scarcity necessarily an important influence on Third World City Growth? A Dynamic General Equilibrium urbanization. The findings suggest that fuel prices have Approach. Princeton, N.J.: Princeton University Press. been less influential in promoting urbanization than the Tolley, G. S., and G. K. Kripalani. 1974. "Stages of Develop- relative price of manufactured goods. Kelley and Wil- ment and Urbanization." In A. A. Brown, J. A. Licori, and liamson concur with Tolley that productivity advances E. Neuberger, eds., Urban and Social Economics in Market that favor modern urban activities have been a potent and Planned Economies. New York: Praeger. influence on urbanization. They see trade policy in in- World Bank. 1984. World Development Report 1984. New dustrial countries and price policies in developing coun- York: Oxford University Press. 2 Urbanization and Economic Development George S. Tolley oncerns about urbanization and economic develop- the question of how the causality works. Economic de- _ ment go hand in hand, for a number of reasons. velopment makes it possible for people to produce more. Beyond the question of whether urbanization is a cause Why should this possibility lead to urbanization? Can or a consequence of development, it is the backdrop for our understanding of the links between development development in almost all nations. Rapid urbanization and urbanization be carried to the point of quantifica- poses complex management problems in housing, trans- tion? What are the main implications of a better under- port, land use, the environment, energy, public services, standing of urbanization? and finance. These problems are responsible in part for In seeking answers to these questions in this chapter interest in whether overurbanization has occurred and it is recognized that the explanation of urbanization is whether more active policies are needed to control and largely an explanation of economic activity. Cities would direct urbanization. To deal adequately with urbaniza- not grow unless there were productive things for people tion issues, an understanding is needed of why the pro- to do in them. cess takes place and what determines its rate, its geo- In the past hundred years enormous growth of cities graphic pattern, and the characteristics of people in occurred in the Western world as people took the fruits cities. of development in the form of products, other than food That there is a close relation between urbanization and fiber, that could be produced more economically in and economic development is evident from experience cities than in the countryside. Rapid urbanization in in most countries that have undergone development. It developing countries today is in some part a result of a also seems clear that urbanization is not a primary cause similar wish to take the fruits of development in the of economic development. form of goods and services suited to urban production. In a very long view of history the role of cities in But more than in the past, urbanization appears to be promoting economic development-through being cen- related to international specialization that is freed from ters of the learning and interactions that lead to in- great dependence on domestic demand. People are pro- novation-would be a concern. In recent history, as ducing goods in the cities of the developing world and communications have been revolutionized, economic trading them in international markets for all manner of development has come more to consist in applying other goods, including food. The links between the com- ideas, no matter where they originate. position of domestic consumption and urbanization It seems reasonable that some, and perhaps most, have been greatly weakened. Actually, the links were urbanization is the result of economic development pro- never exclusive, and they have become progressively less cesses. But accepting a general line of causality from so throughout the world with the increasing importance economic development to urbanization does not answer of international trade to all economies. Domestic and international demands for urban and Note: Kenneth Miranda provided valuable help in the preparation of nonurban goods, as just discussed, are the basis for this chapter. demand influences on urbanization. The ability to pro- 15 16 George S. Tolley duce more urban and nonurban goods acts to increase income and the relative prices of the two types of goods. abundance and is a supply consideration. Population These factors are in turn influenced by supply condi- growth results in growth of human inputs and is like- tions. The total resources of an economy are devoted to wise a supply consideration. Many of the puzzles of either urban or nonurban activities. The amounts of urbanization can be understood by analyzing how these output of the two types of activities depend on how the demand and supply influences interact. The demand and available resources are split between them and on pro- supply influences provide headings under which urban- duction considerations that determine how much of ization is analyzed in this chapter. each type of output can be obtained from a given amount A traditional model of urbanization, appropriate to of resources. closed economies, is considered first. Relations between Over time the total number of persons working in the urban and nonurban productivity, income elasticities of economy may grow, and for both urban and nonurban demand, and population growth as determinants of activities the amount of output that can be obtained for a urbanization are analyzed. Then urbanization is con- given amount of inputs may increase. Given the in- sidered in a more general setting, with international creases in the number of persons, in the total stock of trade. Attempts are made to distinguish between first- capital in the economy, and in productivities, the re- order and second-order effects for countries that are sponses to the demand and supply changes that occur price takers in world markets. One of the conclusions during economic growth will determine changes in the from the resulting model, which is applicable to most allocation of resources between urban and nonurban countries, is that greater urban productivity leads to activities. more urbanization rather than (as in the traditional The view that urbanization accompanies economic model) less. Symmetrically, and again contrary to the development implies that the demand and supply traditional model, greater nonurban productivity leads changes are such that increasing urbanization in fact to less urbanization. results. Why is this, and is it necessarily to be expected? The model is applied to sixty-six developing econo- The explanation of urbanization has been best explored mies, leading to explanations of the degree of urbaniza- for the high-income Western countries which have ex- tion in each economy in terms of growth in urban perienced dramatic declines in the proportion of people productivity, nonurban productivity, and population. working in agriculture and in this sense have become The use of the framework developed in this chapter to the most fully urbanized. The received explanation of further deepen understanding of urbanization in indi- this result centers on one element on the demand side vidual countries is discussed, and some important policy and one on the supply side. implications are noted. Income Elasticities of Demand Demand and Supply Factors The element on the demand side is income elasticity of demand. Because of the low income elasticity of de- A first step toward an understanding of urbanization mand for agricultural products-say, 0.10 to 0.20- is to consider the underlying demand and supply condi- people are viewed as choosing to spend an increasing tions that determine how the resources of an economy proportion of income on urban products as per capita are divided between urban and nonurban activities. A income rises. The low income elasticity of demand for first approximation is to associate nonurban activities agricultural products has thus been cited as one of the with agriculture and urban activities with industry and principal reasons why a society devotes more and more other nonagricultural pursuits. The distinction ignores of its resources to nonagricultural pursuits and thus nonagricultural production by farm households and in- becomes more highly urbanized. Since people must dustry in rural areas, and it begs the question of exactly spend all their income on either urban or nonurban how urban is defined-that is, whether small towns are commodities (where commodities are comprehensively urban or rural and where the dividing line between defined to include both consumption goods and the urban and nonurban phenomena lies. Although the dis- capital goods on which savings are spent), the income tinction is not perfect, the discussion here proceeds on elasticity of demand for agricultural (nonurban) com- the assumption that some degree of approximation is modities can be translated into an income elasticity for inevitable and that nonurban situations in most parts of urban commodities. If all income elasticities are unity, a the world are overwhelmingly agricultural. 1 percent increase in income leads to a 1 percent in- Demand conditions determine how people split their crease in expenditures on every commodity. If the in- income between urban and nonurban products. Impor- come elasticity for some commodities is less than unity, tant factors in the allocation of income are the level of expenditures on them will rise by less than 1 percent and Urbanization and Economic Development 17 the income left over will enable expenditures on other commodities rises or falls during development depends commodities to rise by more than 1 percent. Thus, if the on which effect predominates. income elasticity for nonurban products is less than It is conceivable that the income elasticity for urban unity, the income elasticity for urban products will be commodities is low during the early stages of develop- greater than unity. ment, rises to a peak during a middle period, and falls at The balance of income elasticities above and below the later stages. If so, there will be a tendency for the rate unity depends on the relative amounts of expenditure on of urbanization to start out low, rise to a peak, and fall to each kind of commodity. A well-known condition is that a low rate. The first column of table 2-1 shows various the expenditure-weighted sum of income elasticities values of the income elasticity of nonurban commodities over all commodities is equal to unity. For example, if 20 and the second column shows accompanying percent- percent of expenditures are for nonurban commodities ages of expenditures on those commodities that might and 80 percent for urban commodities, and if the in- be observed. The third column shows the income elas- come elasticity of demand for agricultural products is ticity of demand for urban commodities that results 0.10, then the income elasticity for urban products is from these combinations, as derived from the formula 1.225, implying a 1.225 percent increase in expenditures for -9, above. The entries in the table progress from the on urban products for every 1 percent increase in in- top row, which corresponds to conditions that might be come. To obtain this result, note that the weighted encountered in the initial stages of development, to income elasticity condition is w11¶111 + WaT-a = 1, where intermediate situations (the middle of the table), to the ws are the urban and nonurban expenditure weights advanced development (the last row). and %1 and q, are the income elasticities. The solu- In the first row the income elasticity of demand for tion of the income elasticity for urban products, %, is agricultural products is at its high value, unity. The (1 - waTja)IWu, or, with the numbers in the example, income elasticity of demand for urban products is then (1 - 0.2 * 0.1)/0.8, or 1.225. also unity, no matter what the expenditure fractions are. The example is for a high-income country; a different The income elasticity effect would not lead to any rela- result may be obtained for a low-income country. It is tive urbanization, for it would leave the urban propor- generally thought that the income elasticity demand for tion of the population unchanged. In the second set of agricultural products goes down as income rises. Thus, entries the income elasticity of demand for agricultural for lower-income countries a higher agricultural in- commodities has fallen to 0.8. Some possible values of come elasticity implies a lower income elasticity for the expenditure share accompanying this elasticity are urban commodities and hence, other things being equal, shown. If the share of expenditure on agricultural com- a lower rate of urbanization. A country could still experi- modities remains high (80 percent), the income elastic- ence urbanization, but the lower income elasticity for ity for urban commodities rises to 1.8; it does not rise as urban commodities might contribute to a lower rate of much if the expenditure shares are less. In the next set of urbanization in the early stages of development. entries the income elasticity for agricultural products is In addition to differences in income elasticities (the 0.5. If the expenditure share has fallen by this time to 50 -qs), another difference between high-income and low- percent, the income elasticity of demand for urban com- income countries-and one that operates in the oppo- modities will be 1.5, with higher and lower values for site direction-is the fraction of expenditure devoted to different expenditure shares. The last row corresponds urban and nonurban commodities (the ws). If the frac- tion spent on nonurban commodities is high, the weight on the income elasticity for nonurban commodities will Table 2-1. Determinants of Income Elasticity also be high. At an early stage of development, when per of Demand for Urban Commodities capita income is low, a large fraction of income is likely Income spent to be spent on agricultural commodities. The income Income elasticity of on nonurban Income elasticity of demand, nonurban commodities demand for urban elasticity of demand for agricultural commodities will commodities (percent) commodities not be as far below unity as in a high-income country and, because its weight is greater, it will tend to make 1.0 0-100 1.0 the income elasticity for urban commodities only slight- 0.8 80 1.8 ly above unity. 0.8 50 1.2 As development proceeds, the gradual decline in the 0.8 20 1.05 income elasticity for agricultural commodities will act 0.5 50 1.5 to increase the income elasticity for urban commodities, 0.5 20 1.125 while the decline in the agricultural weight will act to decrease it. Whether the income elasticity for urban 01 20 1.225 18 George S. Tolley to the example originally given for a high-income coun- accompanied by increasing specialization of production. try, with an income elasticity for urban commodities of Indeed, the very possibility of economic development 1.225. has often been based on such specialization. Increasing Table 2-1 suggests that the income elasticity for urban international specialization is part of a wider process of commodities will first rise and then fall during develop- specialization and provides a reason for increasing ur- ment. If the share of expenditure on agricultural com- banization that is related to economic development. modities falls only slowly, the income elasticity for Increasing specialization can occur in agricultural as urban commodities rises longer and further than if the well as in urban production. It is conceivable that some share of expenditures on agriculture falls rapidly. countries could become less rather than more urbanized In addition to demand conditions within the country, during development, although no prominent example the proportion of the population which is urbanized will comes to mind. A large part of the internationally traded be affected by changes in the degree of international portion of agricultural production comes from high- specialization that occur during development. A country income countries, which are already highly urbanized. may develop primarily by increasing its output of manu- Growth of domestic demand and comparative advantage factured goods for export to the rest of the world. If in nonagricultural production are more important in agricultural output does not increase enough to meet explaining urbanization than is specialization in inter- domestic demand, the exports of manufactured goods national agricultural trade. will be used in part to pay for increasing imports of agricultural commodities. As a result, the increase in Su production of urban commodities will be larger than if pply Factors there were no changes in the degree of international When attention turns from the demand to the supply specialization, and a smaller fraction of the increased side, concern shifts from changes in output to changes demand for agricultural commodities will be met from in the numbers of people used to produce the output. If production inside the economy. The income elasticity human inputs were equally substitutable for physical for agricultural commodities must be redefined to per- capital in urban and in nonurban production activities, tain only to the percentage increase in domestic agri- and if the rates of growth of factor productivity in the cultural production associated with a 1 percent rise in two types of activities were the same, changes in output income, and increases in output of urban commodities and changes in numbers of people would go together will be larger than implied by domestic elasticities. exactly, and the demand considerations already dis- Changes in the degree of international specialization cussed would suffice to explain the rate of urbanization may thus accelerate the rate of urbanization during during development. But this is not the case. Conditions development. vary widely, and output per person in production of Changes in international specialization appear to be urban commodities may rise more or less rapidly than the rule and not the exception during development. for nonurban commodities. They have been particularly pronounced in such dramat- Traditional explanations have emphasized rapid tech- ic growth experiences as the "Pacific growth miracles" nological change in agriculture, which leads to more in which Hong Kong, Japan, Korea, and Taiwan all rapid growth of output per agricultural worker as the experienced rapid development simultaneously with in- major supply consideration that acts in conjunction creases in output of nonagricultural goods for export. In with low income elasticity of demand to eject resources these economies and in others with similar, although from agriculture. Several considerations other than less extreme, growth patterns, concern has arisen over technological change, however, affect output per the loss of agricultural self-sufficiency as imports of worker, as will now be discussed. agricultural products rise along with the rising output of Land scarcity is overrated as a factor that could de- the economy. A more positive way of looking at the crease output per worker in agriculture. If output did go decline in agricultural self-sufficiency is to view it as a down, the rate of urbanization would be retarded as manifestation of gains from international specialization. more workers were kept in agriculture to produce a It is more economical to produce manufactured goods given agricultural output. This outcome could occur if and exchange them for agricultural products than to growing population pressed seriously against scarce forgo the additional manufactured output and try to land supplies, but the evidence even for the prime exam- produce agricultural products that can be produced ple of a country that might be in these straits-China- more efficiently elsewhere. is unclear. New land has been coming into production in Although generalizations are hazardous, historically some parts of the world, notably in Latin America. Im- economic development has been almost universally provement of the quality of land through investment Urbanization and Economic Development 19 retards declines in output per worker. International be eliminated in nonurban than in urban areas. Here, specialization has worked against agriculture in many again, countries are not alike. Family and home countries and has eased pressure on land supply. businesses abound in many cities, as do small unviable Agricultural research has effects opposite to those of farming operations in many parts of the world. (Much scarce land supply. An increase in output per worker in campesino agriculture is of this type.) The important agriculture permits a given demand to be met with fewer characteristic of traditional production for the purposes agricultural workers, releases people for urban pursuits, of this discussion is that, in spite of all of its adjustments and increases the rate of urbanization. Even if the in- to relative prices and other changes during develop- crease in productivity as a result of agricultural research ment, it is outcompeted by other forms of production. leads to income growth, workers will be released to During economic development people are released from urban areas if the income elasticity of demand for the the declining traditional forms of production to other products whose output is increased is less than unity, as activities. If they are released from traditional employ- seems likely. It is also true that the Green Revolution ment in urban areas and go into other employment in and other innovations resulting from agricultural re- urban areas, there is no effect on urbanization, but if search have sometimes been more saving of land and they are released from traditional employment in agri- related physical inputs than of workers. The effect of cultural areas where there are no expanding employ- these innovations has in any case been extremely varied ment opportunities, they will seek employment in urban among and within countries, so that it is more than areas, and urbanization will increase. usually difficult to generalize about it. Because of the low amount produced per person in As development proceeds there is a tendency for more traditional production, a reduction in traditional em- of the inputs used in agriculture to be produced in urban ployment leads to increased output per person. Whether areas. This phenomenon is part of the increasing spe- output per person rises more in urban or in nonurban cialization that has already been mentioned and appears areas depends on the rate of reduction in traditional to be speeded up by innovations resulting from agri- production in the two types of areas and on its relative cultural research. Gasoline, feed concentrates, and fer- importance in the areas. tilizers and other nutrients are among the inputs pro- Changes in the quality of human inputs through duced in urban areas that replace inputs formerly education and other forms of human capital formation produced on farms. In addition, home garden plots and deserve mention in relation to urbanization. Urbaniza- home food processing decline. tion and human capital formation both appear to be Successful economic development is characterized by responses to factors that lead to economic development, rising returns to human inputs, resulting in part from rather than urbanization being a primary cause of hu- substitution of capital for labor. These changes may man capital formation or vice versa. Like physical capital affect urban and nonurban production differently. Ur- formation, human capital formation results in increases banization will be favored if the substitution of capital in output per person and allows a given demand to be for labor is higher in nonurban than in urban produc- fulfilled by fewer people. Given the demands for urban tion-another of the circumstances that may differ from and for nonurban commodities, a higher rate of increase one country to another. A large source of the substitu- in human capital per person in nonurban than in urban tions is replacement of labor-intensive human or animal production will be conducive to a high rate of urbaniza- sources of power by machine power, and possibilities for tion, since the demand for nonurban commodities can substitutions abound in nonurban as well as urban pro- be fulfilled by fewer people. The increase in educational duction. (The mechanization of cotton growing in the levels and the growth of skilled and professional catego- United States had wholesale effects on rural-to-urban ries of employment in cities are sometimes thought to migration and profoundly affected American cities.) be necessary conditions for development. Yet farmers Urban production, which uses readily transferable tech- also become better educated. Nonurban production may niques from other countries, may offer greater possibili- use a lower amount of human capital per person at the ties for substitutions, at least in manufacturing indus- early stages of development, and the level may remain tries. But urban production is more intensive than lower through a great part of the development process nonurban production in services, where capital appears even though human capital per person in nonurban to be less substitutable for labor. production is continually rising and may indeed be ris- One of the effects of economic development is the ing at a faster rate than in urban production. The faster gradual elimination of so-called traditional production rise in human capital per person in nonurban than in in both urban and nonurban areas. Urbanization will be urban production leads to the hypothesis that increases speeded if there is more of this traditional production to in nonurban human capital could contribute to a high 20 George S. Tolley rate of urbanization. This effect, however, may be sec- Considerations that affect output per person are ondary to the consideration that both urbanization and brought in by noting that the percentage increase in human capital formation are basically results rather output per capita for any commodity is equal to the than causes of economic development. percentage change in output per person engaged in production of the commodity plus the percentage Population change in the proportion of people in the nation engaged in that production. The second and third conditions are The demand and supply considerations discussed so thus ,u = Tu, + hu for urban commodities andy0= T0 far pertain to the relative rate of urbanization, or urban- + h, for nonurban commodities, where T is percentage ization per capita, and would be sufficient to explain change in output per person engaged in production and urban growth in a country with stable population. Grow- n is percentage change in the proportion of people en- ing total population provides a kicker to both urban and gaged in production. The second and third conditions nonurban population growth. The rapid population follow from differentiation of the identity that output growth rates in some parts of the world mean that much equals number of workers times output per worker. The urbanization comes simply from the fact that there are fourth condition is that any change in the urban pro- more people in the country. portion of the population must be accompanied by an equal and opposite change in the nonurban propor- The Traditional Model of Urbanization tion. In percentage terms the fourth condition is k1 = - (Na IN,1)ia, where N0 and N11 are the numbers of The discussion has considered the role of the income persons in nonurban and urban areas. The four con- elasticities of demand in determining urbanization, the ditions that have been given determine the four vari- factors that affect output per person in each sector, and ables j11, Ya, nu, andha0. The first term on the right-hand the role of total population growth. The following for- side of the above expression for the rate of growth of mula brings those factors together into an expression for urban population is the solution of the set of four condi- understanding their combined influence on the rate of tions for izh, the percentage change in the proportion of urbanization. the population which is urban. As can be seen from equation 2-1, the numerator of (2-1) ( / ) r1TT - )V + N. the term giving the percentage change in the proportion N11(lu"laN1/N0) + 1 urban is equal to the negative of the rate of growth of Since the urban population N,, equals the proportion output per person in production of urban commodities urban n,, times total population N (or, N,, = n 1N), the (that is, an increase in urban productivity, acting alone, percentage change in the urban population N,, is the leads to a decrease in urban population because the percentage change in proportion urban h41 plus the per- demand for urban commodities can be met by fewer centage change in total populationN(or,&N = flu + N), people) plus the ratio of urban to nonurban income which is the basis of equation 2-1. The variable on the elasticities multiplied by the rate of growth of nonurban left-hand side, N1, is the annual percentage rate of output per person (growth in nonurban output per per- growth of the urban population. The first term on the son releases persons for urban activities, and the magni- right-hand side is the percentage rate of change in the tude of the effect depends on the income elasticities). proportion of the population which is urban. It includes The denominator-one plus the product of the ratio of all the considerations discussed earlier except the effect income elasticities and the ratio of the numbers of peo- of total population growth of the country, which is given ple in urban and in nonurban areas-arises because of by the second term on the right-hand side, N. the interactions of the various effects. Note that the ratio of the percentage increases in The second term in equation 2-1 isN, the percentage output per capita of urban and nonurban commodities rate of growth of total population of the country. Its will be in the ratio of their income elasticities. (Recall presence indicates that when changes in the proportion that the income elasticities are redefined to take account of the population that is urban are taken into account, of international trade effects.) This first condition can be the urban population varies in direct proportion to the writteny,/ya = iu/ %, where j11 is the percentage rate number of people in the country. of growth of urban output per capita (4u = Y11/N, or From the formula it is clear that rates of urbanization urban output divided by the country's total population) may differ with conditions in particular countries. As an and r, is the trade-adjusted income elasticity of demand example consider a country, still at an early stage of for urban commodities. Similar definitions apply to Ya development, in which per capita income is rising some- and 9, where the subscript a stands for nonurban what rapidly. If the growth is oriented toward growth of commodities. manufacturing products that are exported to world mar- Urbanization and Economic Development 21 kets. the ratio of the increase in urban and in nonurban that developing countries are increasingly getting out of output, % I' , will be rather high-say, 4. Suppose that the agricultural business and into urban-based produc- one-third of the population is in urban areas and two- tion. thirds is in rural areas, so that the value forNu/N, is 0.5. Although the model of the preceding section allows Output per worker in urban areas is assumed to be rising for the effects of trade in the specification of income at2 percent ayear (Tu = 0.02).As a result of progress in elasticities, it is not fully consistent with the general a commercialized part of agriculture and the decline of reasons for rising urban population connected with spe- subsistence agriculture, output per worker in agricul- cialization in production noted above. In equation 2-1 a ture is also assumed to rise at the rate of 2 percent a year rise in agricultural productivity will increase urbaniza- (Ta = 0.02). It is assumed that the total population of tion because a given demand for agricultural products the country is growing at 1.5 percentayear(N = 0.015). can be satisfied by fewer people and the excess people The values are substituted into the formula for the yearly move to the cities. In like fashion, a rise in urban pro- percentage growth in the urban population. ductivity acts to decr:ase urbanization, since fewer peo- 4 0.02 - 0.02 ple are needed to satisfy the demand for urban products. (2-2) N 4 = 4 0.5 + I + 0.015 0.035. The amounts of labor required for production in both sectors change according to a life of their own, and That is, the annual growth rate for the urban population amounts of output are not affected by productivity is 3.5 percent. changes. Examples of much more rapid or of slower urbaniza- Yet one might think that if a sector became more tion can be constructed by varying the values. The pur- efficient, it would increase its production, not decrease pose of this section, however, has not been to explain the it. An increase in efficiency lowers the price at which a rate of urbanization for any one country but to begin to given quantity can be produced. If the price elasticity of suggest the nature of the underlying changes in demand demand is greater than zero, a lowering of the price will and supply that determine urbanization. increase the amount that can be sold. A production response is to be expected. Indeed, if the responses to changes in production take A More General Explanation the form of adjustments in foreign exports and imports, prices are effectively set in world markets where de-- The discussion points to generalizations about why mands are highly elastic and where in many cases prices worldwide urbanization is occurring. One reason is that can be taken as given. Even within an economy, a more incomes are rising in many places, albeit in varying adequate view of the differences between urban and degrees. The higher income elasticity of demand for rural production will recognize that there is not a strict urban than for nonurban commodities is a fundamental dichotomy in the types of goods produced. Aside from reason why urbanization can be expected to accompany local or nontraded goods in each sector, there are goods economic development throughout the world. that may be produced in either sector and that can be Another reason for world urbanization is more closely traded between the two sectors. There may be little farmn related to the nature of the economic development that output in urban areas (although chickens and garden is occurring. For much of the world economic develop- plots can be significant in cities), but factory production ment entails a transfer of techniques of production, and and home production of urban-type goods abound in by and large the techniques are more transferable for rural areas, and trade between sectors in processed foods urban than for agricultural commodities. Techniques of and other commodities is commonplace. For most of the agricultural production are much more tied to local land world it may be that most consumption in either rural or and climate conditions. Efforts are being made to de- urban areas is of goods produced within each area and velop and adapt agricultural techniques to suit the that trade in marketed surpluses of rural and urban unique conditions of each area of each country, but the goods is confined to a small part of output, much of pace, at best, is slow. Moreover, agriculture, as an inher- which can actually be produced in either sector. ent user of location-specific resources, is subject to These considerations of the way trade occurs suggest intrinisic comparative advantages. Increasing spe- that it may be more reasonable to take prices facing rural cialization of agricultural production among major re- and urban areas as given rather than to take relative gions of the world may be in store. quantity growth as given, as in the model presented Urban production, on the other hand, is not nearly as above. If, as a first approximation, prices are taken as location-bound. Factories, machines, and electronics given, productivity changes will have the opposite effect parts tend to be physically similar the world over in their from that implied by the model. A rise in productivity in requirements for productive inputs. It is not surprising urban areas enhances cities' ability to compete in inter- 22 George S. Tolley national markets and to compete with production in degree of mobility that leads to rough equalization of rural areas and thus leads to an increase in urban pro- opportunities for real earnings between sectors, the lat- duction, not a decrease. An analogous result is obtained ter appears far more defensible. One is likely to find that, in nonurban areas, with the implication that rising historically, real eamings of comparable labor in urban nonurban productivity will retard urbanization. The re- and rural areas have changed by about the same amount, sults are consistent with the hypothesis that urbaniza- and the degree of net and gross migration observed tion results from changing advantages in production. between urban and rural areas is so great as to belie the Given that a rise in productivity leads to an expansion notion that responses do not occur. The single most rather than a contraction of sector output, the question important distinguishing feature in the relations be- becomes how great the expansion will be. Since relative tween the urban and rural parts of an economy may be quantities of output change are endogenous rather than that they are connected by migration. If one is not to exogenous, a careful distinction is needed between base an explanation of urbanization on a growing dispar- changes in average labor productivity and changes in ity between real eamings of comparable labor in urban marginal labor productivity as affected by diminishing and in rural areas, it is reasonable to suppose that there returns to employment in a sector. If there were no is sufficient labor mobility to roughly equalize the urban diminishing returns as employment increased in a sec- and rural changes in marginal productivity that occur tor, the assumption of given output prices (suggested during economic development. The implications of here as a first approximation) would imply that, except lesser degrees of mobility could be considered, but this by accident, one sector would have a higher marginal can be done more easily and with more insight after the productivity of labor than the other. Migration would first approximation is explored. then lead to all the employment being in one sector or In distinction to the variable T that was used above to the other-one sector would go to zero and the other refer to average labor productivity, let A denote the would have all the employment. influence of nonlabor inputs on the amount of output To avoid this unrealistic solution the assumption of (taken to be synonymous with income produced). For given prices might be dropped, leading to a solution nonurban production, the availability of land may partly governed by changing prices of urban goods in relation determineA. Port facilities and other natural limitations to those of nonurban goods. But the idea that urbaniza- could play a similar role in urban production. For both tion is primarily explained by massive changes in prices urban and nonurban production, private capital and the of urban goods in relation to those of nonurban goods- amounts of various types of infrastructure influence A. in particular, by increases in relative prices of urban With labor N measured in terms of numbers of people or goods-is not appealing. This type of change does not hours, A is influenced by the productive quality per appear to be a pervasive world phenomenon. Endoge- person, which in turn is determined in part by human nous relative prices can be introduced, but it would be capital investment. For both urban and nonurban pro- best to do so only after a situation of given relative prices duction, perhaps the greatest influence onA is technical has been more fully explored. knowledge used in production, which depends on en- Another way to avoid the unrealistic solution of hav- couragement of or impediments to international trans- ing all employment end up in one sector or the other fer of technology, entrepreneurship, incentive struc- would be to base a solution on impediments to labor tures, and other policy and institutional considerations. migration. Then a rise in, say, urban compared with To obtain a formulation consistent with growth nonurban labor productivity would increase the attrac- source analysis, take time derivatives of the production tiveness of urban employment. Imperfect labor mobility relation Y = AN6 and express as yearly percentage would lead to a rising gap in real incomes between urban changes. Then growth in output will equal the percent- and rural sectors that might be reduced only gradually age growth in nonlabor influences, A, plus the product over time. One problem with this solution is that it does of elasticity of output with respect to labor, e, and the not avoid the outcome that all employment ends up in percentage change in labor. For urban production, one sector; it only delays that outcome. Eventually, over (2-3) y + e generations, if not earlier, people all migrate to the cities aU in response to relative differences in earnings opportuni- and for nonurban production, ties. The only reason for not doing so would be nonpe- (2-4) A, + N. cuniary attachments to rural life. Basing an explanation of urbanization on permanent. nonpecuniary attach- The elasticity of output, e, is almost certainly between ments to a way of life is at least as unappealing as basing zero and one and is closer to one than to zero. If it were it on changing relative output prices of the sectors. one or greater, the implication would be that-other If one has to choose between no labor mobility and a inputs being held constant-output could be increased Urbanization and Economic Development 23 in equal or greater proportion to increases in labor. above, E = wnIpy, can be rearranged as an expression Under the assumption of fixed proportions the elasticity for the wage in a sector, w = epyIn. With the use of would be zero. That assumption, however, ignores the rates of change for the two sectors, myriad of possibilities for drawing on a given amount of nonlabor inputs to produce varying physical quantities (2-5) Wu = Pu + Yu - and qualities of outputs in modern factory production, and office work, retailing, personal services, and traditional production as the amount of labor is varied. The relation (2-6) Wa = Pa + Ya - Na between average labor productivity T in the preceding assuming that e is a stable parameter (e = 0). The Ps in section and the influence of nonlabor inputs A in this these conditions refer to prices of output produced; the section can be seen by combining the conditions T = relevant deflator that determines labor supply response YIL and Y = AL' to obtain T = ALE- 1. If e is close to one is an index of prices of goods consumed. The level of and if nonlabor influences A are more predominant as prices may differ greatly between urban and rural areas. causes of change in average labor productivity than the If prices move proportionately over time and relative elasticity effect of changes in labor inputs LE -1, then T real earnings in urban and rural areas remain roughly and A will move in a similar fashion. the same, the rate of growth of wages in the two sectors To proceed to the rest of the model in this section, will be similar, which yields the condition Wu, = Wa. suppose that labor is paid a wage w that is equal to its The right-hand sides of equations 2-5 and 2-6 can then marginal product m multiplied by price output p (or, m be set equal to each other. = wlp). The elasticity of output (the marginal product If the approximation is retained that changes in prices m multiplied by the amount of labor n divided by output of urban-produced goods relative to those of rural goods y) is seen to be equal to wage payments wn divided by the will not in the first instance be used as an explanation of value of output py. This implies that e in equations 2-3 the rate of urbanization, 1% = Pa. Then the condition and 2-4 equals the share of product paid to labor, wn Ipy. obtained from setting the right-hand sides of equations The order of magnitude of e as labor's share can in 2-5 and 2-6 equal to each other is principle be observed rather readily. Although the elas- - ticity of output may be different in urban and in nonur- (2-7) Nu = - Na. ban production, in this example it is taken to be the This condition has the strong empirical implication that- same. Not to do so would imply that major reliance is the excess of the rate of income growth over the rate of placed on differences in factor shares in explaining ur- employment growth will be the same in urban and in banization. nonurban areas. Put another way, average labor produc- The rate of urbanization may indeed be affected by tivity will change by the same proportion in both sec- urban-rural differences in elasticity of output as well as tors. by changes in elasticities in particular countries. There Income produced in the two sectors may change by are no compelling studies that indicate systematic differ- greatly different amounts-that is, Yu, and Ya, may be ences and changes in factor shares, and the shares do very different. The differential changes in income pro- seem in fact to be fairly stable. The formula for elasticity duced are a result of differences in growth of nonlabor of output followed here therefore seems appropriate for sources of income, which lead to corresponding differ- an overview of major factors that explain rate of urban- ences in growth of demand for labor. For example, if ization. nonlabor sources of income growth increase more rap- Given observed values of e, equations 2-3 and 2-4 can idly in urban than in rural areas, the demand for labor be used in connection with measurements of income will increase more in urban than in rural areas. The growth Yand population growthNto obtain estimates of increase in nonlabor sources of income and the associ- the growth of nonlabor influences, A. When either equa- ated increase in labor inputs will raise income by 1, and tion 2-3 or equation 2-4 is solved forA,A is found to be Y employment by some lesser amount, N . Meanwhile, minus eN. The estimate of growth of nonlabor sources because of the slower growth of nonlabor income of income in a sector is the observed income growth less sources in rural production, Yk will be smaller than Yk that part of income growth attributable to labor, which and NQ will be correspondingly smaller than N& (owing in turn is equal to labor's elasticity of output multiplied to a lesser growth in demand for labor in rural areas). If by the population growth of the sector. factor shares and relative wages and prices are not to An essential feature of the model is that labor will change, the rise in ratio of income to employment must adjust to keep the relative earnings of urban and rural be the same in the two sectors. employment in line. To incorporate this feature, the How well does the condition stated in equation 2-7, condition pertaining to the elasticity of output derived that average labor productivity will change by the same 24 George S. Tolley proportion in urban and rural employment, fit the facts? errors or differences in the way variables are calculated For sixty-six low- and middle-income economies for from available data. The uncertainty of output and em- which the requisite income and population figures are ployment figures is well known, and the reasons for available to study growth rates for 1960-80, the mean of differing changes in measured employment and output the excess of the proportionate change in average urban connected with changing market orientation of produc- labor productivity over average rural productivity is tion during development have often been discussed. In - 0.0001. The standard error of this mean is 0.0029. In general, more production may pass through the market- these comparisons real nonagricultural and agricultural place as economic development proceeds over a twenty- incomes are used as surrogates for urban and rural year period, but because of lags in data collection, statis- incomes. It is assumed that employment is roughly tics may capture these changes only imperfectly. proportional to population, so that urban and rural Nonzero error of fit may also occur because assump- populations can be used as surrogates for employment. tions of the model that are justified as first approxima- Alternatively, still with nonagricultural and agri- tion may not be exactly fulfilled. But the small size of the cultural incomes as surrogates for urban and rural in- errors suggests that the principal factors are indeed comes, the labor force in agriculture can be used as a captured by the model. In view of all the possible reasons surrogate for rural employment and the nonagricultural for existence of error, it is fair to say that the data fit the labor force as a surrogate for urban employment. Or, model well. more simply, the comparison may be viewed as an ap- It was specified that growth in output can be repre- plication of the model to agricultural versus nonagri- sented as the sum of labor and nonlabor effects in each cultural employment rather than to urban versus rural sector in equations 2-3 and 2-4. Condition 2-7 (for employment. If agricultural and nonagricultural labor changes in employment relative to output in each sector forces are used, rather than rural and urban popula- as migration occurs in response to changes in demand) tions, the mean of the excess of the proportionate was then derived. The model is completed by using the change in average nonagricultural labor productivity condition discussed above-that the total amount of over that of average agricultural productivity is 0.0024 labor used in the economy is the sum of urban and and the standard error is 0.0029. nonurban labor, or N = N11 + N,. The equation is To further test the extent to which equation 2-7 is differentiated with respect to time and divided by the satisfied, if the error attributable to the fact that original values of the variables. The annual proportion- observed changes in average productivity do not exactly ate rate of growth of the total population is expressed as fulfill the equation is small, most of the observed the product of the urban share of population and annual changes in output and employment will be explained by proportionate rate of growth of the urban population, the model. The small size of the error in relation to total plus the product of the nonurban share and the annual income or employment change is an indication of the proportionate rate of growth of the nonurban popula- extent to which the hypothesis of equal proportionate tion; that is, changes ih average labor productivity is fulfilled. 2-8' N F N F N An error-of-fit measure, calculated for each economy, ( u a a- is the proportionate annual change in average labor The four-equation model given by formulas 2-3, 2-4, productivity as a fraction of the mean of annual propor- 2-7, and 2-8 determines the four rates of growth that tionate urban and rural population growth. The median pertain to outpyt and employment in the two sectors, of the error-of-fit measures over all sixty-six economies that is, Y, Ya, N, and NA . To solve for the variable of is 0.030. Two-thirds of the measures fall between interest-urban population growth, Nu-equations 2-3 - 0.735 and 0.826. When nonagricultural and agri- and 2-4 are substituted into equation 2-7 to eliminate cultural labor forces are used instead of urban and rural the Ys. The result is a relation between N1u and N1 that populations, the median error-of-fit measure for the depends on the rates of growth of nonlabor sources of sixty-six economies is 0.067. Two-thirds of these mea- income in the two sectors. The relation can be written Na sures fall between - 0.981 and 0.828. = N11 + (A,, - A,,) / (1 - E), indicating that urban and The fact that most of the error-of-fit measures are rural populations move proportionately except as they substantially less than 1 indicates that most of the are influenced by differential changes in growth of non- observed variation in output and employment is associ- labor sources of income, which have a multiplier effect ated with growth of nonlabor income sources and the of 1 / (1 - e) because under the model enough labor hypothesized proportionate changes in labor and that must be hired to drive the return to labor down to the relatively little change is left to be explained by errors or same level in each sector. The result for Na is inserted departures from the hypotheses of the model. The exis- into condition 2-8 for total employment and the equa- tence of nonzero error of fit may be partly a result of tion is solved for Nk: Urbanization and Economic Development 25 (F. _ F. mates of growth of urban and rural income, which may (2-9) N11 + 10 -Ael - ela be subject to great measurement errors and possible - inconsistencies in definition. To minimize the effects of The solution of the model forN11 given by equation 2-9 measurement problems, the procedure used here is to provides the basis for decomposing the rate of growth of average the changes in income in the two sectors and the urban population into three factors. The first is total thus obtain income measures that conform to the pro- population growth of the nation, N, which has a propor- ductivity condition. Specifically, the average excess of tionate effect on urban population growth. The second is the rate of growth of income in each sector over the rate an urban productivity factor that stems from growth of of population growth in each sector yields an estimate of nonlabor sources of urban production. The positive the common rise in average labor productivity in the two effect of increases in nonlabor sources of urban produc- sectors. This estimate made on the basis of the observed tion is enhanced by the multiplier 1 / (1 - e) that is average for the two sectors can be added to population connected with keeping earnings growth the same in the growth in each sector to obtain the measures of income two sectors, and the magnitude of the effect is also growth in the sectors that would prevail if the model proportional to the fraction of the population in agricul- fitted the real situation precisely. After this attempt to ture. A given change in the absolute level of employment purge the income measurements of error and to elimi- as labor is reallocated will have a greater proportionate nate the effects those errors would spuriously introduce effect on urban population the smaller is the existing into estimates of the factors that affect urbanization, fraction of the population that is urban (which is to say, estimates of the As may be obtained as before by sub- the larger the fraction that is nonurban). The third tracting the product of labor elasticity of output and the factor into which urban population growth is decom- amount of labor from the income values that conform to posed is a rural productivity factor which stems from the average labor productivity condition. growth of nonlabor sources of rural income. It is exactly Table 2-2 presents the results obtained when the mod- analogous to the urban productivity factor, but it acts on el is applied to sixty-six low- and middle-income econo- urban population growth with a negative sign, since in mies, using urban population statistics from the World this model a rise in rural productivity increases rural Bank and an estimate of labor's share Fof 0.7. Column 1, population. the annual proportionate growth of urban population in each economy from 1960 to 1980, is the variable to be Application to the Sample explained. Column 2 is the average fraction of the Appicafton to the Sample population that is urban for 1960-80. This fraction was Measures of the rate of population growth N to be calculated to conform to equation 2-8, which relates the applied to equation 2-9 are readily available. The fraction rates of urban and rural population growth to the rate of of the population that is not urban, Fa, is available for growth of the total population. The fraction of the many countries but may be subject to more error than population that is rural, F,, is 1 - F,,. Arrangement of total population. Systematic measures of labor's elastic- equation 2-8 as an expression for F gives as the fraction ity of output F that are comparable among sectors and of the population that is urban F, = (N - Na) I (Nu1 - nations are apparently not available, but impressions Na), the ratio between the excess of total population about general orders of magnitude can be obtained. growth over nonurban population growth and the excess Although the growth in nonlabor sources of income (the of urban over nonurban population growth. This mea- As) cannot be directly observed, first-cut measurements sure is quite close to the simple average of the 1960 and can be obtained with the use of equations 2-3 and 2-4. 1980 values for the fraction of the population that is That is, when either equation is solved for A, the con- urban, and in all cases it lies between those values. The tribution of growth in nonlabor sources of income to methodjust described for calculatingF11 in conformance observed income growth is equal to the observed income with equation 2-8, along with the method described growth Y minus the product of elasticity of output of above for calculating the growth of nonlabor income labor F and the growth rate of labor N. (That is, the sources A to conform to equations 2-3 and 2-4, ensures nonlabor source of income growth is total income an internally consistent system of accounting in which growth minus the contribution of labor to income the factors that explain urban population growth exactly growth.) add up to urban population growth. Although other The calculations just described give only first-cut or, procedures for defining variables might have given at most, suggestive measurements of the As because about the same results, the procedures used here have they ignore the stipulation of the model that average the advantage of eliminating idiosyncracies in definition labor productivity rises at the same rate of growth in as a perceived reason for differences in rate of urbaniza- both sectors. The measures are dependent on the esti- tion. The analytic factors that explain urban population 26 George S. Tolley Table 2-2. Urban Population Change, Sixty-six Economies, 1960-80 Components of change Anmual pro- Annual pro- in urban population Annual pro- portionate portionate portionate Urban growth in growth in Annual pro- Urban pro- Nonurban change share urban non- nonurban portionate ductivity productivity in urban of popula- labor nonlabor change in factory factor, Error of fit, population, tion, 1980, sources of sources of total popu- F,A&I -FaAal (Au - A,)! Nu Fu income, Au income, A,, lation, N (1 - e) (I- E) '/2 (u +X N) Economy (1) (2) (3) (4) (5) (6) (7) (8) Bangladesh 0.066 0.076 0.006 - 0.007 0.027 0.017 0.022 - 0.413 Ethiopia 0.070 0.099 0.016 0.001 0.024 0.048 -0.002 -0.171 Mali 0.054 0.151 0.016 0.005 0.024 0.045 -0.015 0.196 Somalia 0.052 0.231 -0.025 - 0.036 0.024 -0.065 0.093 1.527 Chad 0.067 0.118 -0.010 -0.026 0.020 -0.030 0.077 -1.101 Burma 0.039 0.228 0.019 0.012 0.022 0.048 -0.030 - 1.193 Burkina Faso 0.045 0.068 -0.001 -0.010 0.016 -0.002 0.032 0.828 India 0.032 0.199 0.019 0.015 0.022 0.051 -0.041 0.567 Malawi 0.070 0.062 0.028 0.016 0.029 0.089 - 0.049 - 0.249 Rwanda 0.062 0.029 0.018 0.007 0.027 0.059 - 0.024 - 0.096 Sri Lanka 0.041 0.223 0.019 0.012 0.021 0.050 -0.030 -0.315 Guinea 0.061 0.141 0.013 0.002 0.029 0.038 -0.006 -0.190 Zaire 0.060 0.242 - 0.006 -0.021 0.023 - 0.016 0.053 -0.176 Niger 0.070 0.091 -0.001 - 0.014 0.031 -0.003 0.042 -0.554 Benin 0.044 0.119 0.007 0.001 0.027 0.021 -0.004 0.633 Pakistan 0.041 0.249 0.030 0.025 0.029 0.074 -0.062 0.442 Tanzania 0.072 0.080 0.023 0.009 0.028 0.070 - 0.026 0.004 Central African Republic 0.051 0.315 0.007 -0.005 0.022 0.017 0.012 - 0.747 Haiti 0.055 0.247 0.004 -0.012 0.016 0.009 0.030 - 1.060 Mauritania 0.128 0.102 0.009 - 0.025 0.026 0.026 0.076 - 0.313 Lesotho 0.067 0.033 0.039 0.025 0.021 0.126 - 0.080 0.969 Uganda 0.077 0.080 0.006 - 0.008 0.034 0.018 0.026 - 0.724 Sudan 0.070 0.166 0.003 - 0.013 0.024 0.009 0.037 -0.027 Togo 0.053 0.132 0.034 0.025 0.026 0.098 - 0.071 0.671 Kenya 0.070 0.101 0.017 0.005 0.035 0.051 - 0.016 - 0.246 Senegal 0.029 0.240 0.010 0.008 0.025 0.025 - 0.021 - 0.538 Indonesia 0.035 0.174 0.037 0.031 0.020 0.101 -0.087 0.826 Egypt 0.031 0.415 0.037 0.033 0.022 0.073 -0.065 1.046 Ghana 0.049 0.292 -0.001 -0.011 0.026 -0.002 0.025 -0.735 Honduras 0.054 0.292 0.017 0.007 0.032 0.040 - 0.017 -0OA73 Zambia 0.054 0.301 0.019 0.008 0.029 0.045 - 0.020 - 1.449 Thailand 0.033 0.135 0.046 0.045 0.029 0.133 -0.129 1.101 Bolivia 0.042 0.284 0.025 0.018 0.026 0.059 - 0.043 - 0.030 Philippines 0.038 0.330 0.032 0.028 0.029 0.071 - 0.062 0.009 Congo 0.028 0.350 0.004 0.001 0.023 0.008 - 0.002 1.138 Nigeria 0.047 0.163 0.024 0.016 0.025 0.066 - 0.045 1.500 El Salvador 0.036 0.395 0.022 0.020 0.033 0.043 -0.040 0.495 Morocco 0.044 0.348 0.021 0.013 0.027 0.045 - 0.028 0.260 Peru 0.047 0.567 0.027 0.014 0.028 0.039 -0.020 -0.149 Cote d'lvoire 0.080 0.278 0.032 0.018 0.045 0.077 -0.042 0.009 Nicaragua 0.043 0.470 0.041 0.034 0.031 0.073 - 0.060 -0.494 Colombia 0.045 0.593 0.040 0.026 0.026 0.054 -0.035 - 1.148 Paraguay 0.031 0.375 0.034 0.032 0.027 0.071 - 0.067 0.077 Dominican Republic 0.055 0.402 0.032 0.019 0.029 0.064 - 0.037 0.044 Guatemala 0.037 0.360 0.035 0.031 0.028 0.075 - 0.066 - 0.057 Syrian Arab Republic 0.047 0.434 0.048 0.040 0.032 0.091 -0.076 -0.384 Malaysia 0.036 0.270 0.042 0.039 0.028 0.103 - 0.095 0.652 Jamaica 0.034 0.419 0.015 0.005 0.015 0.029 - 0.010 -1.283 Korea, Republic of 0.056 0.410 0.064 0.047 0.022 0.125 -0.092 0.334 Urbanization and Economic Development 27 Table 2-2 (continued) Components of change Annual pro- Annual pro- in urban population Annual pro- portionate portionate portionate Urban growth in growth in Annual pro- Urban pro- Nonurban change share urban non- nonurban portionate ductivity productivity in urban of popula- labor nonlabor change in factor, factor, Error of fit, population, tion, 1980, sources of sources of total popu- FaA]u - FaAal (Au - a,)/ N,, F,, income, A. incomne, A. lation, N (1 - e) (1 - e) 1N2 W +a) Economy (1) (2) (3) (4) (5) (6) (7) (8) Turkey 0.047 0.383 0.036 0.026 0.025 0.075 -0.053 0.301 Algeria 0.065 0.452 0.002 - 0.017 0.030 0.004 0.031 - 0.195 Mexico 0.046 0.592 0.029 0.018 0.033 0.039 -0.025 0.122 Panama 0.044 0.475 0.034 0.026 0.030 0.060 -0.046 0.012 Taiwan 0.038 0.680 0.057 0.044 0.024 0.061 -0.046 1.408 Chile 0.028 0.749 0.027 0.016 0.019 0.022 -0.014 -2.719 Costa Rica 0.038 0.400 0.033 0.039 0.030 0.067 -0.059 0.276 Brazil 0.046 0.556 0.042 0.030 0.028 0.062 - 0.045 0.260 Uruguay 0.010 0.821 0.016 0.011 0.008 0.009 -0.007 -5.082 Argentina 0.019 0.782 0.029 0.022 0.014 0.021 -0.016 -0.914 Portugal 0.018 0.269 0.029 0.023 0.003 0.070 -0.055 5.940 Yugoslavia 0.030 0.348 0.042 0.033 0.010 0.091 - 0.071 0.430 Venezuela 0.044 0.756 0.042 0.029 0.034 0.035 - 0.024 - 1.436 Hong Kong 0.023 0.895 0.044 0.042 0.023 0.015 -0.015 2.877 Greece 0.024 0.526 0.054 0.042 0.006 0.085 -0.066 -0.601 Spain 0.024 0.658 0.048 0.037 0.011 0.055 -0.042 0.894 Israel 0.039 0.836 0.050 0.037 0.031 0.027 - 0.020 - 0.642 Note: Economies are listed in increasing order of 1978 per capita income. growth in fact add up exactly to measured urban popula- Column 8 presents the error-of-fit measure for each tion growth. economy. As discussed earlier, that measure is the dif- Columns 3 and 4 present measures of the average ference between urban and nonurban average productiv- annual rate of growth of the sources of nonlabor urban ity change (based on observed income) divided by the and nonurban income A, with the use of the purged average employment change in the two sectors. A posi- income values, as described above. tive sign indicates a greater urban than nonurban aver- The key columns are 5, 6, and 7. They show the age increase in labor productivity, using observed mea- decomposition of the annual proportionate urban sures of income; a negative sign indicates the reverse. A population growth given in column 1 into the three value of less than one in absolute terms indicates that terms in equation 2-9. Thus, column 5 is the annual relatively more of the employment changes are associ- proportionate overall population growth of the econ- ated with the factors explained by the model than with omy, which makes a direct and equal contribution to failure to satisfy exactly the proportionate productivity urban population growth. Column 6 is the urban pro- condition. The effect of the error is smaller the closer is ductivity factor, which gives the effect of nonlabor urban the value to zero. income source growth on the rate of urban population Column 5, proportionate total population growth, growth. Column 6 is column 3 multiplied by the nonur- shows, as expected, that total population growth ban fraction of the population (calculable from the table accounts for a rate of urbanization that is generally on as 1 minus the urban fraction in column 2) and divided the order of 2 percent a year (0.020, as a decimal frac- by the share of income that accrues to factors of produc- tion, as expressed in the table). It ranges up to 3 percent tion other than labor, 1 - e (in this case, 1 - 0.7, or in some cases and for certain economies, such as Yugo- 0.3). Column 7, the nonurban productivity factor, is slavia, is very low. The direct effects of population calculated from column 4 in the same way as the urban growth thus cannot account for the high rates of urban- productivity factor is calculated from column 3, except ization-up to 7 percent-which are found in the table. that the sign is changed, since nonurban productivity Population growth could also cause pressure on limited acts in the opposite direction from urban productivity as agricultural land supplies and force growing numbers of an influence on the rate of urbanization. people to seek work in urban areas. This effect would be 28 George S. Tolley manifested in a resource scarcity situation in which to be associated with low rates of growth of nonlabor supplies of nonlabor resources were more expansible in rural income sources does not prove that land scarcity urban than in rural areas. If nonlabor resources were as has been a highly important reason for rapid urbaniza- scarce in urban as in rural areas, in the sense of having tion. A more likely explanation of the finding may be the an elasticity of supply that is zero or equally small, general sluggishness of agriculture in the twenty-five nonlabor inputs would tend to increase by similar rela- economies and a general failure to advance knowledge tive amounts in both rural and urban areas and there or introduce new techniques-in short, a failure of non- would be little impetus to urbanization. In a situation in land conventional and nonconventional inputs to in- which limited supplies of land provided the major im- crease, even though land supply may have been elastic. petus to urbanization, one would expect to see limited Lack of inputs other than land may have acted to depress increases in nonlabor sources of income growth in rural the marginal productivity of labor and hence the de- areas, coupled with larger increases in labor inputs in mand for labor in rural areas. If so, it is a lack of effective urban areas as the supply of urban sites with associated economic development in agriculture and not a lack of infrastructure and other urban facilities expanded to land. that helped to foster high urbanization rates among accommodate people displaced from agriculture. These the countries that experienced low rates of growth of endogenous increases in urban nonlabor inputs would nonlabor rural income sources. be greater than the increases in agricultural nonlabor The interpretation of events in terms of a more gener- inputs, but if the only important impetus to urbaniza- al economic development perspective that goes beyond tion were displacement of agricultural labor, the in- preoccupation with the role of agricultural land is creases in urban nonlabor inputs still might not be strengthened by considering additional results in table great, since in this depiction of the world more vigor- 2-2. One of the striking relations in the table is the ous positive impetuses to economic development are positive association between growth rates of urban and lacking. rural nonlabor sources of income, as revealed by looking One might suppose that economies with less than 1 down columns 3 and 4. The two series have a simple percent a year growth in agricultural nonlabor inputs correlation coefficient (R2 ) of 0.964. Two reasons may may have a land scarcity situation. In table 2-2 twenty- be noted for the positive association. First, the nonlabor five of the sixty-six economies have growth of nonlabor income sources include capital, which is mobile between sources of nonurban income of less than 1 percent a year sectors. Increases in the capital stock of an economy will (or, as a decimal fraction, 0.010) as indicated in column tend to be allocated among sectors so as to equalize the 4. Of these, ten experienced negative rates of change in marginal returns to capital and impart changes in the nonlabor sources of income, which indicates retrogres- same direction to nonlabor income sources in the dif- sion and other special factors not explained by the usual ferent sectors. Second, and perhaps more important, the land scarcity situation. Note that land scarcity does not nonlabor income sources reflect more general economic imply negative growth of nonlabor sources of income development changes, in addition to conventional in- but rather no growth or a very small increase. The puts, that act to increase the productivity of the conven- economies with low and negative growth for nonlabor tional inputs. There is some tendency for the effects of sources of nonurban income tend to be among those economic development changes to be pervasive; for ex- with the lowest incomes, as indicated by their early ample, the influence of research, infrastructure, incen- appearance in the table. (Economies are ranked by 1978 tive structures, and institutions tends to be felt through- per capita income.) It is also true that these economies out an economy. Just as there is a tendency to equalize generally have high urbanization rates. In the five returns to conventional inputs among sectors, there is lowest-income economies (Bangladesh, Ethiopia, Mali, some tendency to equalize returns to deeper economic Somalia, and Chad) rates of change for nonlabor sources development efforts, even though the latter tendency of rural income are negative or less than 1 percent a may be less perfect. The fact that the economies in table year. Each had a rate of growth of urban population of 2-2 which have low rates of growth of nonlabor rural over 5 percent a year (0.050, as a decimal). Of the income sources also tend to have low rates of growth of twenty-five economies with low or negative growth of nonlabor urban income sources suggests that general nonlabor sources of rural income, only three-Senegal, economic development differences, and not primarily Congo and Jamaica-had an urban population growth of agricultural land, are at work in explaining those econo- less than 4 percent a year. By contrast, in only twenty of mies' high rates of urbanization. the forty-one economies where growth of nonlabor An ancillary finding in table 2-2 is that the growth in sources of rural income was greater than 1 percent were nonlabor sources of income tends to rise with the per annual rates of urban population growth more than 4 capita income of the economy (which is to say, it rises as percent. one looks down the table). The simple R2 between the The apparent tendency for high rates of urbanization income rank of an economy (as indicated by its order in Urbanization and Economic Development 29 the table) and the urban nonlabor income source growth population analysis in which labor share, E, is 0.5 in- rate in column 3 is 0.658. The simpleR2 between income stead of 0.7; and two labor force analyses that use nonag- rank and the rural nonlabor income source growth rate ricultural labor force as a surrogate for urban labor in column 4 is 0.642. The finding reflects the tendency force, with e at 0.7 in one case and 0.5 in the other. The among these economies for the more affluent among results were essentially similar to those presented here. them to gain in relation to the less affluent. The existence of many variations from the general To return to factors that affect the rate of urbaniza- patterns that have been discussed indicates that condi- tion, it is noteworthy that when columns 3 and 4 of table tions in individual economies -play a part in urbaniza- 2-2 are compared, for every one of the sixty-six econ- tion. India and Burma are among the economies which omies the growth rate of urban nonlabor income stand out particularly as exceptions to some of the pat- sources is greater than that for rural nonlabor income terns in table 2-2. Both rank low in per capita income sources. This finding is consistent with the hypothesis and have high nonurban proportions of population, yet noted earlier, that the techniques of urban production unlike most other economies with these characteristics are easier to transfer among economies than those of they have relatively low urbanization rates. Growth rates agricultural production. Since an excess of urban over of urban and rural nonlabor income sources are largely rural nonlabor income source growth acts to foster offsetting in these two countries. The size of the Indian urbanization in this model, the rate of urban population economy might influence the result, but this considera- growth is greater than the rate of overall population tion would not apply to Burma. growth for every one of the sixty-six economies. India and Burma are only two examples that indicate A final consideration in interpreting table 2-2 is the the usefulness of more analytic work for understanding role of the fraction of the population which is urban. If patterns of urbanization. For any one of the economies, this fraction is small, any differential change between more detailed analysis could be undertaken, with more urban and rural nonlabor income source growth has a refined identification of the factors that affect urbaniza- magnified effect on urban population. The reason is that tion as an object. a given absolute change has a greater percentage effect Further comparative analytic work to contrast and on a small number than on a large number. This effect is understand differences among regions and different reflected in equation 2-9, in which the coefficient Fa / types of economies is recommended to strengthen the (1 - e) multiplies the growth rate of nonlabor income knowledge base for prediction and policy concerned with source to arrive at the urban and rural productivity economic development. Further work should give atten- factors. Thus, economies with large nonurban fractions tion to differences in industrial structure and to consid- in their populations tend to have greater urbanization erations that affect trade. Greater detail on types of rates. Since these economies generally have low per goods could be included, as could allowance for price capita incomes, they are the ones that have predomi- effects, age composition, education, dependency, and nantly experienced the highest urbanization rates. other considerations bearing on work participation. The Progressing down the table to increasingly higher- assumption that the number of workers in urban areas income economies, one finds that the difference be- goes hand in hand with the population of these areas tween urban and rural nonlabor income source growth would be modified. stays about the same or becomes larger, but the con- tributions of the urban and rural productivity factors decrease because of the declining value of the multiplier Fa I (1 - E). Since Fa = 1 - Fu, the multiplier will be Conclusion greater than one if the fraction of the population that is urban is greater than the share of income that accrues to This chapter suggests that the degree to which a labor. The condition would be fulfilled for all but a country succeeds in fostering growth of urban produc- handful of the economies, regardless of the value of e, tivity is likely to be the major determinant of its urban- but the multiplier falls more and more toward one ization. Population growth will continue to be a source toward the bottom of the table. The rates of growth of of both urban and nonurban growth. Countries that lag urban population decline until they are not much in growth of agricultural productivity will experience greater than the rate of overall population growth. This added urbanization pressures. is shown in equation 2-9, which gives an analytic expla- As important as these general findings is the useful- nation for the descriptive finding of the association be- ness of the framework of this chapter as a basis for tween the fraction that is urban and the growth rate of quantifying the experience of individual countries. Pre- urban population. diction and planning are enhanced, and the economic The analysis presented here of factors that affect ur- factors that influence the effectiveness of policies con- banization was carried out for three other cases: a cerned with urbanization can be better understood. 30 George S. Tolley Bibliography Development and Cultural Change, vol. 3, no. 1 (October), pp. 4-5. Armstrong, Harvey, and Jim Taylor. 1978.RegionalEconomic . 1957. "Urbanization and Economic Growth in Asia." Policy and Its Analysis. Deddington, Oxford, U.K.: Philip Economic Development and Cultural Change, vol. 6, no. 1 Allan. (October), pp. 42-54. Arriaga, Edward. 1970. "A New Approach to the Measurement Kelley, Allen C., and Jeffrey G. Williamson, 1979. Modeling of Urbanization." Economic Development and Cultural Urbanization andEconomic Growth. Working Paper 79-81, Change, vol. 18, no. 2 (January), pp. 206-18. Laxenburg, Austria: International Institute for Applied Sys- Berry, Brian J. L. 1961. "City Size Distributions and Economic tems Analysis. September. Development." Economic Development and Cultural Lasvwn, J. R. 1973. "Urbanization and Development-The Change, vol. 9, no. 4, pt. 1 (July), pp. 573-88. Temporal Interaction Between Geographical and Sectoral Bourne, L. S., and J. W. Simmons, eds. 1978. Systems of Cities: Cluster." Urban Studies, vol. 10, no. 2 (June), pp. 163-88. Readings on Structure, Growth, and Policy. New York: Linn, J. F. 1979. "Automotive Taxation in the Cities of Develop- Oxford University Press. ing Countries." Nagarlok (January). Breese, Gerald. 1966. Urbanization in Newly Developing Lipton, Michael. 1977. Why PoorPeople Stay Poor: UrbanBias Countries. Englewood Cliffs, N.J.: Prentice-Hall. in World Development. Cambridge, Mass.: Harvard Uni- Chapin, Stuart F., Jr. 1971. "Selected Theories of Urban versity Press. Growth andStructure." In L. S.Bourne, ed.,InternalStruc- Miller, J. C. 1979. "Regional Development: A Review of the ture of the City: Readings on Space and Environment. New State of the Art." AID 20523. Agency for International De- York: Oxford University Press. velopment, Washington, D.C. Processed. Colosio, Donaldo. 1979. Urbanization andEconomic Develop- Mills, Edwin. 1971. "City Sizes in Developing Economies." In ment in Mexico. Working Paper 79-19. Laxenburg, Austria: Raanan Weitz, ed., Urbanization and the Developing Coun- International Institute for Applied Systems Analysis. tries. Report on the Sixth Rehovot Conference. Praeger Currie, L. 1975. "The Interrelation of Urban and National Special Studies in International Economics and Develop- Economic Planning." Urban Studies, vol. 12, no. 1 (Febru- ment. New York: Praeger. ary), pp. 37-46. _ . 1972. 'Welfare Aspects of National Policy toward City Davis, Kingsley. 1965. "The Urbanization of the Human Popu- Sizes." Urban Studies, vol. 9, no. 1 (February), pp. 117-24. lation." Scientific American, vol. 213 (September), pp. 28, Pascal, Anthony H. 1971. "Urban Growth and Development: 40. Discussion." American Economic Review, vol. 61, no. 2 - 1971. "The Role of Urbanization in the Development (May), pp. 357-59. Process." In Raanan Weitz, ed., Urbanization and the De- Renaud, Bertrand. 1981. National Urbanization Policies in veloping Countries. Report on the Sixth Rehovot Confer- Developing Countries. New York: Oxford University Press. ence. Praeger Special Studies in International Economics Reynolds, L. G. 1975. "China as a Less Developed Economy." and Development. New York: Praeger. American Economic Review, vol. 65, no. 3 (June), pp. 418- Felstehausen, H. 1971. "Planning Problems in Improving Co- 28. lombian Roads and Highways." Land Economics, vol. 47, Richardson, Harry W. 1977. City Size and National Spatial no. 1 (February), pp. 1-13. Strategies in Developing Countries. World Bank Staff Work- Friedmann, J., and C. Weaver. 1979. Territory and Function. ing Paper 252. Washington, D.C. University of Califomia Press. Rondinelly, Dennis A., and Kenneth Ruddle. 1978. Urbaniza- Hansen, Niles M. 1969. "Regional Development and the Rural tion and Rural Development: A Spatial Policy for Equitable Poor." Journal ofHuman Resources, vol. 4, no. 2 (Spring), Growth. New York: Praeger. pp. 205-14. Tolley, George S. 1974. "The Welfare Economics of City Big- Hansen, Niles M., ed. 1976. Public Policy and Regional Eco- ness." Journal of Urban Economics, vol. 1, no. 3, pp. 324- nomicDevelopment. TheExperience ofNine Western Coun- 45. tries. Cambridge, Mass.: Ballinger. Tolley, G. S., and G. D. Gwyer. 1967. "Intemational Trade in Harberger, C. 1970. "Economic Policy Problems in Latin Agricultural Products in Relation to Economic Develop- America: A Review." Journal of Political Economy, vol. 78, ment." In Herman M. Southworth and Bruce F. Johnston, no. 4, pt. 2 (January-August), pp. 1007-16. eds., Agricultural Development and Economic Growth. Henderson, J. V. 1980. "A Framework for International Com- New York: Cornell University Press. Pp. 403-47. parisons of Systems of Cities." Urban and Regional Report Tolley, G. S., and G. K. Kripalani. 1974. "Stages of Develop- 80-3, World Bank, Washington, D.C. Processed. ment and Urbanization." In A. A. Brown, J. A. Licori, and Holleb, D. B. 1976. "Mao's Legacy for Revolutionary Urban E. Neuberger, eds., Urban and Social Economics in Market Development." Working Paper. University of Chicago, Cen- and Planned Economies. New York: Praeger. ter for Urban Studies. December. Tolley, G. S., and S. Smidt. 1964. "Agriculture and the Secular Hoselitz, Bert. 1954. "Introduction; The Role of Cities in Position of the U.S. Economy."Econometrica, vol. 32, no.4 Economic Development and Cultural Change." Economic (October), pp. 554-75. Urbanization and Economic Development 31 Tolley, G. S., and Barton Smith. 1975. "Social Costs and the tives for Spatial Intervention for Developing Countries." Rural-Urban Balance." In Earl 0. Heady and Larry R. Whit- Working Paper. Chicago: University of Chicago. ing, eds., Externalities in the Transformation of Agricul- Townroe, Peter M. 1979. Employment Decentralization Pol- ture: Distribution of Benefits and Costs from Development. icy Instruments for Large Cities in Less Developed Coun- Ames, Iowa: Iowa State University Press. tries. Vol. 10, pt. 2, in D. R. Diamond and J. McLoughlin, Tolley, George S., Philip E. Graves, and John L. Gardner, eds. eds., Progress in Planning. Elmsford, N.Y.: Pergamon. 1979. Urban Growth Policy in a Market Economy. New Wheaton, W. C., and H. Shishido. 1979. "Urban Concentra- York: Academic Press. tion, Agglomeration Economies, and the Level of Economic Tolley, G. S., R. J. Krumm, and P. Graves. "Design of Incen- Development." Working Paper. Cambridge, Mass.: MIT. 3 What Drives City Growth in the Developing World? Allen C. Kelley and Jeffrey G. Williamson C ities are capital-intensive, a stylized fact which has developing-country urbanization in the recent past can C led many pessimists to assert that current rates of be traced to the availability of cheap energy, technolog- city growth in the developing world cannot be sustained. ical diffusion that favors modem urban sectors, heavy To make matters worse, cities are stocked with public capital inflows, world trade liberalization, a drift toward capital at low or zero user charges, and in that sense they domestic price distortions that favor city output, and are too capital-intensive. Since the social costs of in- unusually rapid population growth. That these condi- migration thus exceed private costs, the number and tions have begun to show signs of changing suggests sizes of cities turn out to be excessive. Furthermore, the that the developing world may have overurbanized in increasing scarcity of urban land encourages the sub- the recent past. Urban growth may be expected to de- stitution of capital for land and increases capital inten- celerate over the next two decades under certain condi- sity. When the economy can no longer finance such tions: a recurrence of fuel scarcity, technological regres- urban costs-when it fails, for instance, to cope with sion in modern sectors in the face of a productivity social overhead and housing requirements-the process slowdown in industrial countries, diminished capital of urbanization is retarded. transfers because of economic austerity in industrial No doubt the qualitative analysis is correct: cities are countries, a retreat toward protectionism in industrial capital-intensive and have a voracious appetite for sav- countries, and a decrease in population growth. ings and accumulation. But is the quantitative empha- In short, the pessimists offer three sources for an sis-the idea that cities are "too big"-warranted? incipient urban crisis in the developing world: a savings It is also often alleged that rapid city growth in the constraint which will bridle the growth of capital- developing countries represents a disequilibrium, that intensive cities; a labor market disequilibrium which has these countries have "overurbanized, " and that a painful made overurbanization a temporary but serious problem structural adjustment will eventually be required. Ana- of overshoot; and the disappearance of unusual external lysts of urban problems in the developing world point conditions which were favorable to urban growth in the out that migrants are attracted to the cities in the hope past. Although the pessimists have established a plausi- that they (or their children) will be selected for training ble case, no one to our knowledge has offered a quantita- and employment in protected, high-wage sectors. tive assessment of the importance of these forces over Apparently, new in-migrants are willing to accept under- the past two decades. Without such an assessment, de- employment in low-wage traditional service sectors bate over future trends in developing countries will be while waiting. Eventually, however, social discontent is dominated by allegation and anecdotal evidence. It is likely to erupt. Are labor markets in these cities really in our view that the debate can be better informed by the serious disequilibrium, and is the developing world application of a general equilibrium model of develop- overurbanized as a consequence? ment that includes some of the costs of urbanization, It is often alleged that the unusually high rates of so that "natural limits" to urban growth can be evalu- 32 What Drives City Growth in the Developing World? 33 ated and the impact of changing economic and demo- 1984b), only a suggestive outline is presented here. The graphic conditions assessed (see Kelley and Williamson model is in the neoclassical general equilibrium tradi- 1984a, b). tion. Prices of outputs and inputs are completely flexi- ble, and most are endogenously determined; firms are driven by profit maximization; households are driven by The Limits on City Growth utility maximization; and even government demand de- cisions obey well-defined rules from consumer demand The limits on urban growth are set, on the one hand, theory. Mobility of capital and labor is constrained to by urban costs that affect migration decisions and, on reflect the institutional realities of factor markets in the other, by rising urban investment requirements that developing countries, but economic agents are moti- compete with productive capital accumulation. Urban vated to search for the optimal sectoral and spatial use of land constraints raise rents, increase living costs in resources. urban areas relative to rural areas, and thus inhibit The model has eight sectors (table 3-1). Tradables and migration into cities. To the extent that rising rents and nontradables-the latter include various location- urban disamenities are caused by high densities, crowd- specific services-are distinguished. This is not the first ing, and other manifestations of inelastic urban land multisectoral model to recognize nontradables, but it is supply, city rents reflect the quality of urban life as well the first spatial development model to stress the impor- as living costs. To evaluate the impact of urban land tance of nontradables as an influence on spatial cost-of- constraints on city rents, a general equilibrium model is living differentials, on migration behavior, and thus on needed which, at the minimum, admits housing service the rate of urban growth. activities and confronts equilibrium land use issues. The model is savings-driven, and the aggregate sav- Furthermore, a variety of urban land requirements must ings pool is generated endogenously from three sources: be included-residential squatter settlements, factory retained after-tax corporate and enterprise profits, gov- sites, land use for public social overhead, and luxury ernment savings, and household savings. This savings housing sites. pool is allocated competitively and endogenously to The housing and social overhead investment require- three uses: investment in physical capital (productive ments of city growth must also be analyzed. Such "un- investment), investment in human capital (training), productive" urban investment requirements (first ana- and investment in housing (unproductive investment). lyzed by Coale and Hoover 1958) compete directly with It should be emphasized that these three modes of ac- "productive" capital accumulation and may check urban cumulation are competitive and are determined endoge- growth. If such overhead investment is forgone, housing nously; that is, investment in skills (training) takes place costs rise, the quality of urban services falls, and migra- up to the point at which rates of return are equated to tion to cities is discouraged. Thus, in addition to the the economywide rate on physical capital accumulation. effect of the rise in the relative cost of living in the city, Physical capital goods are allocated across the three the rise in unproductive investment requirements in capital-using sectors so as to minimize differentials in cities may lower the rate of productive capital accumula- after-tax rates of return. Dwelling investment utilizes tion and job creation and set a limit to urban growth. household savings only up to the point at which rates of Modem urban sectors also tend to be relatively inten- return equal the economywide rate on physical capital sive in skills, in intermediate inputs, and in such im- accumulation. Of course, there are institutional and ported inputs as energy. To the extent that cities are technological features which seriously restrict the abil- energy-intensive, fuel scarcity can limit urban growth. If ity of the economy to equate rates of return at the capital and skills are complements and different labor margin. Any of the three dwelling markets (rural, urban skills are poor substitutes for each other, rapid rates of squatter settlements, and formal urban housing) may be urban capital accumulation imply increasing demands starved for funds, since the absence of an intersectoral for skilled labor, which can constrain capacity expan- mortgage market may leave housing investment re- sion, retard growth of employment, and limit urban quirements in excess demand. The immobility of sector- growth. An effort to relax the labor constraint through specific capital stocks makes it likely that current invest- skill accumulation is likely to compete with productive ment allocations are insufficient to equalize sectoral urban capital accumulation and itself constitute a limit rates of return. Indeed, the larger are housing require- on urban growth. ments, the smaller is the residual pool available for This chapter uses a computable general equilibrium physical capital accumulation and the more likely it is model to analyze past, present, and future growth of that current investment allocations are insufficient to cities in the developing world. Since the model has equalize sectoral rates of return. Furthermore, firms' appeared elsewhere (Kelley and Williamson 1980, 1982, demands for skills may remain unsatisfied if the stock of Table 3-1. Sector Characteristics of the Kelley-Williamson Model U.N. ISIC Determination Tradability Production inputs Production Sector Location counterpart of market price characteristics Primary Intermediate fimction form Manufacturing and Urban Manufacturing, mining Exogenous Traded intemation- Capital, skills, Imported raw mate- Nested CES, constant mining (M) ally and interre- labor rials and fuels, A,KS returns gionally Modem capital- Urban Electricity, water, and Endogenous Traded interre- Capital, skills, Imported raw mate- Nested CES, constant intensive services gas, banking, public gionally labor rials and fuels, A,M returns (KS) administration, trade, commerce, construc- tion Informal urban ser- Urban Personal services, some Endogenous Not traded Labor None Cobb-Douglas dimin- vices (US) trade and commerce ishing returns c,, Low-quality (squatter) Urban Dwellings (rent and im- Endogenous, Not traded Dwellings, land None Cobb-Douglas con- urban housing puted ownership) owner-occupier stant returns services (H, US) shadow price High-quality urban Urban Dwellings (rent and im- Endogenous, Not traded Dwellings, land None Cobb-Douglas con- housing services puted ownership) owner-occupier stant retums (H, KS) shadow price Agriculture (A) Rural Agriculture, livestock, Exogenous Traded intemation- Capital, land, Imported raw mate- Cobb-Douglas con- forestry, fishing, ally and interre- labor rials and fuels, M, stant retums hunting gionally KS Informal rural services Rural Personal services, some Endogenous Not traded Labor None Cobb-Douglas dimin- (RS) trade and commerce ishing returns Rural housing services Rural Dwellings (rent and im- Endogenous, Not traded Dwellings None Leontieff (H, RS) puted ownership) owner-occupier shadow price Sources: Kelley and Williamson (1980, 1982, 1984b). What Drives City Growth in the Developing World? 35 potential trainables is insufficient to meet desired levels Macroeconomic Indicators of investment in training. In short, capital market dis- equilibrium may well be a chronic attribute of the econ- Aggregate Growth. As set out in table 3-2, the model omy. generates an annual average growth rate of gross domes- Finally, some exogenous variables that help drive the tic product (GDP) in constant prices of 6.26 percent for economy over time are alleged to have influenced city 1960-73, and the growth rate of GDP per worker averages econmy ver imeare lleed t hae inlueced ity 3.58 percent. A significant trend acceleration is in- growth. These variables are the nominal value of foreign dicated; A gnifrom an acceleratiof 5s92 capital and aid (F) available each year to help finance the percent in 1960-65 to 6.60 percent in 1968-73. That development effort and forestall balance of payments prentins to 6.60 cento the at problems; the total unskilled labor force (L), which is these predictions conform fairly closely to the actual determined by earlier demographic events; the sectoral pre-oPEc history of the RDE group helps to establish the rates of change in total factor productivity, which favor model's plausibility at the most aggregate level. modern sectors and are labor saving; prices of imported Unbalanced Growth and Industrialization. The raw materials and fuels (Pz), which are influenced by model also replicates the RDE history of industrialization actions of OPEC and by other world market conditions; mode clselicateseen in dustheliaton and the terms of trade between primary exportables (PA) quite closely, as can be seen in table 3-3. The M-goods and manufactured importables (PM), which are twisted sector underwent relatively rapid growth, from 15.9 by domestic price ipolante Political economy of percent of GDP at factor cost (current prices) in 1960 to protectionist or liberalization winds in the industrial 20.8 in 1973; the model generates almost identical nations. trends. The service sector underwent a more modest nations. permit generalizationabouturbanizationinthe rise, from 47.5 to 50.6 percent of GDP at factor cost, and To permit generalization about urbanization in the the model captures these trends too. The rapid decline in developing world, the representative developing econ- the relative importance of agriculture-from 36.6 to omy (RDE) has been employed as the data base on which 28.6 percent-is also fithfully replicated in the simula- the model is estimated and validated. The RDE is a fic- tion. It is satisfying that the model so closely replicates tional economy that embodies the experience of devel- tiontitative that the unbalanced rowtes oping economies on three continents since 1960. Three the quantitative record of the unbalanced growth of criteria were used to select the RDE group: availability of sectoral output. data beginning in the early 1960s, low per capita income but some successful growth, and close conformity to the InvestmentAllocatson and Sources of inance. Gross model's most important assumptions. (The last-named domestic investment as a share in GDP increased mark- criterion implied the rejection of economies which have edly between 1960 and 1973, from 15.59 to 19.46 per- been heavily dependent on foreign capital and on min- cent (Kelley and Williamson 1984b, table 3.10, p. 89). beeral hexports suchpasnuel) T fore ital eondoms whin- Total investment shares, including investment in train- eral exports such as fuel.) The forty economies which ing, also drift upward in the model, a prediction which met the criteria' seem to exhibit the same average attn- conventional national accounts cannot document. butes as the far larger group analyzed by Chenery and Although the aggregate investment share increased over Syrquin (1975). The RdE group contains eleven of the the period, the model suggests that the distribution of world's twelve largest developing economies and covers investment was remarkably stable. Investment alloca- more than 80 percent of the population of the develop- tion by sector and type is poorly documented for the pre-OPEC period, but the model predicts that the dis- tribution of investment by sector and type must have been quite stable over the thirteen years. The sources of Predicting the Past Table 3-2. Annual Average Growth Rate, 1960-73 Of the about one hundred endogenous variables (percent) generated by the model, three-urbanization, city GDP GDP per worker growth, and rural-urban migration-form the core of (constant prices) (constant prices) this chapter. The illustrations below show how the Period Model Actual Model Actual model replicates past trends for these and other selected variables, both to throw light on past trends and to 1960-65 5.92 5.78 3.19 3.24 validate the use of the model for making predictions. 1968-73 6.60 6.12 3.98 3.58 Much of the initial discussion focuses on 1960-73, but 1960-73 6.26 5.80 3.58 3.26 the late 1970s are also discussed. Source: Kelley and Williamson (1984b), table 3.5, p. 84. 36 Allen C. Kelley and Jeffrey G. Williamson Table 3-3. Output Mix, 1960-73 (percent of CDP) 1960 1970 1973 Sector Model Actual Model Actual Model Actual Agriculture (A) 36.6 36.6 30.4 30.9 28.2 28.6 Manufacturing and mining (M) 15.9 15.9 19.3 19.2 20.9 20.8 Modern capital-intensive services (KS) 29.6 29.6 30.4 n.a. 30.8 n.a. Informal urban services (US) 6.2 6.2 6.9 n.a. 7.2 n.a. Informal rural services (RS) 3.2 3.2 3.1 n.a. 2.8 n.a. Housing, all sectors (H) 8.5 8.5 9.9 n.a. 10.1 n.a. All services combined (KS + US + RS + H) 47.5 47.5 50.3 49.9 50.9 50.6 n.a. Not available. Source: Kelley and Williamson (1984b), table 3.7, p. 86. saving, however, did not show the same stability (Kelley Urban Indicators and Williamson 1984b, table 3.10, p. 89).2 The share of Migration, Urbanization, and City Growth. Table gross domestic investment financed by public saving 3-4 documents four key aspects of urban development: rose to 34 percent. The model predicts a comparable rise the share of the population that is urban, city growth to 33.7 percent, with both shares rising by about 5 rates, net rural out-migration rates, and net urban in- percentage points. The fall in the private share of finance migration rates. The rise in urban shares provides some was attributable not so much to lagging household sav- evidence of accelerating rates of urbanization and con- ing as to a decline in the share of gross corporate saving forms to the pre-inflection-point phase along logistic in gross domestic investment, from 31.7 to 26.6 percent. urbanization curves that is commonly found in develop- IncomelInequality and Wage Patterns. Size distribu- ing-country time series (Preston 1979; Ledent 1980). In tions for the model and for an average of eighteen of the the first five years of the simulation (1960-65) urbaniza- RoE for which dataexist are given in Kelleyand William- tion levels rise by 3.18 percentage points, while in the RDE or hic daa eistaregive inKeley nd illam- last five years (1968-73) they rise by 5.38 percentage son (1984b, table 3.12, p. 91). In spite of well-known las. Similars city growthes rise by 5.38 cont limitations, the top 5 percent and the bottom 20 percent shares of national income exhibit some striking trends. Contrary to conventional wisdom, an unambiguous rise in inequality is not confirmed. The bottom 20 percent Table 3-4. Migration, Urbanization, suffered an erosion in income shares during the period, and City Growth, 1960-73, Model Predictions but the top 5 percent underwent an even more dramatic Net rural Net urban erosion. Clearly, the middle classes flourished at the Percent City out-migration in-migration expense of both the very rich and the very poor. Those Year urban growth rate rate rate are exactly the conditions under which Lorenz curves 1960 32.60 - - - intersect. Inequality inferences are then impossible un- 1961 33.55 5.56 1.41 2.91 less value weights are imposed on social classes. The 1962 34.30 4.87 1.13 2.24 model exhibits the same behavior: a 3.1 percent fall in 1963 34.73 3.85 0.65 1.25 the share in national income of the top 5 percent is 1964 35.25 4.13 0.81 1.51 predicted, compared with the RDE history of a fall of 2.6 1965 35.78 4.11 0.82 1.50 percent. Similarly, while the bottom 20 percent found 1966 36.49 4.60 1.10 1.98 its share declining by 0.3 percent, the model predicts a 1967 37.23 4.61 1.16 2.01 fall of 0.2 percent. The group in the 60-90 percent range 1969 38.95 4.91 1.425 2.3217 increased its share by 3.7 percent, which implies that the 1970 39.93 5.15 1.60 2.51 source of the inequality lies with increased inequality of 1971 40.96 5.22 1.71 2.58 earnings. Those in the 60-90 percent range in the model 1972 42.15 5.56 2.02 2.91 are urban skilled and unskilled in the formal sectors. In 1973 43.45 5.72 2.21 3.07 short, the model generates wage inequality and skill -Not applicable. scarcity. Source: Kelley and Williamson (1984b), table 3.13, p. 93. What Drives City Growth in the Developing World? 37 forming to the trends reported in the World Tables 1976 Table 3-6. Urban Land Use, Density, and Land (World Bank 1976, table 2, "Social Indicators") between Scarcity, Model Predictions, 1960, 1970, and 1973 1960 and 1970. Rural-urban migration rates also rise. Variable 1960 1970 1973 Table 3-5 presents model predictions and actual ex- perience. City growth rates in the RDE group were 4.60 Share of urban land in squatter percent a year during the 1960s; the model predicts a - settlements (percent) 43.0 52.2 53.9 rate of 4.67 percent. The model predicts a rural out- ,rban land density (persons . area) migration rate of about 1.1 percent a year, and the High-quality housing areas 100.0 152.0 172.7 historical rate ranged between 1 and 1.2 percent; the Squatter settlements 100.0 121.4 141.7 predicted urban in-migration rate is 2.1 percent a year, Shadow site rent on urban land and the historical rate ranged between 1.8 and 2.3 per- (1960 = 100) 100.0 183.3 233.3 cent. Finally, 45 percent of the increase in city popula- Shadow price on urban land tion is accounted for by in-migration in the model; this (1960 = 100) 100.0 195.2 239.7 figure falls between Preston's (1979) estimate of 39 per- Source: Kelley and Williamson (1984b), table 3.14, p. 95. cent (based on twenty-nine developing countries) and Keyfitz's regional estimate of 49 percent (1980, p. 151). Housing Scarcity and Gost-of-Living Differentials. In It is also close to the 42 percent figure for developing addition to land scarcity, excess demand for housing countries reported by Linn for 1970-75 (1979, p. 73). units in the short run and rising costs of housing con- Urban Land Use, Density, and Land Scarcity. Table struction in the long run can inflate the cost of city life, 3.6 reports indicators of urban lad ue an. as summarized in table 3-7. Excess housing demand 3-6 shareportsindicato of urban land devot qustera scity arises because capital market segmentation excludes The share of urban land devoted to squatter settlements itretrl(otae on o osn netet rises sharply (Mohan 1979; Beier and others 1975). mtersectoral (mortgage) loans forhousing Investment, Competition for land use generates sharply rising land and each socioeconomic class must rely on its own scarcity; the shadow site rent on urban land almost internally generated saving to meet housing investment. doubles between 1960 and 1970 and in 1973 reaches a Table 3-7 reports that the saving constraint is never level about 2.3 times that of 1960. Urban land values binding for rural households or for skilled workers and (deflated by the general price level) surge; the index rises higher-income households, but for squatter settle- from a base of 100 in 1960 to 195.2 in 1970 and 239 7 in ments-the faster growing sector-the saving con- from aTbase ofe100 in 1960n to195.2aind 1970uan 239.7 in straint is bindingvery early in the period and thus excess 1973. These tn i urbantlan uly a raid short-run demand for new low-quality dwelling units is annual growth of 7 percent-exactly the sharp rises gnrtd xesdmn n nraigubnln which were singled out at the U.N. Habitat Conference generated. Excess demand and mcreasmg urban land in 1976 (United Nations 1976). Urban densities rise scarcity ensure a rapid upward drift in (shadow) rents in everywhere in the model, but they rise most dramati- cally in "luxury" housing, where the relative scarcity of Table 3-7. Housing Scarcity and Cost-of-Living land compared with structures rises most sharply, en- Differentials, Model Predictions, 1960, 1970, couraging land saving and greater density. and 1973 Variable 1960 1970 1973 Table 3-5. Predicted and Actual Migration, Excess housing investment Urbanization, and City Growth, Averages, 1960-70 Urban high-quality housing -3.77 -2.55 -2.67 (percent) _________________________________________ Urban squatter settlements -0.02 0.37 0.28 Variable Predicted Actual Rural housing -0.29 -0.58 -0.88 Annual city growth (compounded) 4.67 4.60 Housing rent index Total increase in urban share of Urban high-quality housing 1.00 1.35 1.46 population 7.33 5.30 Urban squatter settlements 1.00 1.72 1.90 Annual increase in urban share of Rural housing 1.00 1.38 1.45 population 0.73 0.53 Urban squatter . rural 1.00 1.25 1.31 Net in-migrant share of increase in Cost-of-living index urban population 45.0 39.3-49.0 Urban - rural 1.00 1.09 1.12 Net in-migration rate 2.09 1.81-2.26 Net out-migration rate 1.10 0.97-1.21 a. [Desired housing investment demand - household saving] - desired housing investment demand. Source: Kelley and Williamson (1984b), table 3.13, p. 93. Source: Kelley and Williamson (1984b), table 3.15, p. 96.- 38 Allen C. Kelley and Jeffrey G. Williamson urban squatter settlements, and the cost-of-living dif- In table 3-8 disembodied TFPG (the Ajs, in rates of ferential rises as a result. change) in agriculture and manufacturing exert a much greater impact on urbanization than do productivity improvements in the service sectors. The price variables Understanding City Growth: PA and Pm are exogenously determined and are fixed by Some Major Influences invoking the small-country case of infinite price elastic- ity. Services are nontradables with price elasticities of Which exogenous variables have had the greatest im- demand sufficiently low that the productivity-induced pact on city growth in developing countries? Which are declines in service sector prices (PKs, PUS, and PRS, in least likely to account for future growth? Answers to response to AKS, Aus, and ARS, respectively) imply sta- these questions can be divided into three parts: the size ble marginal value products and trivial employment and of the past and future changes in the exogenous variable city growth effects. Although productivity advance in thought to influence endogenous rates of city growth; manufacturing is an important potential determinant of the short-run comparative static impact of that exoge- urbanization, rapid productivity advance in agriculture nous variable; and the long-run forces set in motion by tends to forestall out-migration to the city. This result is the short-run comparative static impact. This section in contrast to the closed dual economy model in which focuses on short-run comparative static elasticities and productivity advance in agriculture meets with demand explores the impact of some key macroeconomic events absorption problems, declining farm terms of trade, and on urban growth: unbalanced productivity advance, thus a labor surplus which out-migrates to glut urban world market conditions and price policy, accumula- labor markets. tion, demographic change, and land scarcity. Unbalanced rates of TFPG that favor manufacturing are Each of these short-run elasticities reflects the full likely to have been a key determinant of rapid in- general equilibrium impact of the exogenous variable in migration and city growth since the late 1950s. Not only question, based on the initial conditions in 1970. Labor are the comparative static elasticities in table 3-8 consis- markets adjust through migration, and urban land mar- tent with that position, but limited evidence suggests kets seek an optimal land use solution, but capital mar- that annual rates of TFPG in manufacturing have been kets are severely constrained in the short-run analysis: relatively high in most successful developing econo- old capital cannot migrate, and new capital goods and mies. newly trained skilled workers are not added to capacity. Although technological advance tends to be lower in Investment responses are also ignored in the short-run the service sectors, especially in the informal service analysis: recent historical experience with sectoral in- sectors, table 3-8 suggests that rapid TFPG in those sec- vestment allocation is assumed to guide entrepreneurs tors has had little impact on urbanization experience for who are slow to adjust to the new, unexpected, and the demand elasticity reasons already offered. If rapid shock-distorted rates of retum. TFPG matters little to urbanization even in the modem capital-intensive and skill-intensive KS sector, lagging Unbalanced Productivity Advance productivity advance in the service sectors would also matter little. It is believed that urban social overhead is If output demand is relatively price inelastic, sectoral crucial to the profitability and viability of urban firms. In total factor productivity growth (TFPG) for a sector tends our model, KS activities such as transport, communica- to generate a relative price decline rather than an elastic tions, and electricity have that role. Table 3-8 confirms supply response. Thus, the rise in the marginal physical that productivity advance augments KS services sup- productivity of factors used in a technologically dynamic plied in short-run equilibrium (QKs), but because sector will be partially offset by the decline in price, so almost all of these productivity gains are passed on to that marginal value products rise by less, and shifts of users elsewhere in the economy (PKS declines), employ- resources, including labor, to the technologically dy- ment in KS itself changes little. Final-demand custom- namic sector are minimized. If, on average, urban sec- ers of KS services, who tend to be the urban rich, gain. A tors tend to have relatively high rates of TFPG, and if the major user of KS services as intermediate inputs is demand for urban output is price elastic, final demand manufacturing, and employment rises there. The net shifts toward the dynamic urban sectors, the derived effect on urbanization is, however, indirect and small. demand for urban employment is augmented, and the city grows. The higher are price elasticities of demand Short-Run Constraints on City Growth for urban output, the greater is the city growth attribut- able to unbalanced productivity advance that favors the Productivity advances that favor modern sectors fos- modem sectors. ter urbanization, but in the short run the city growth What Drives City Growth in the Developing World? 39 Table 3-8. Comparative Static Impacts of Unbalanced Productivity Advance on City Growth in Developing Countries (1970 Elasticities) Tradable commodities Nontradable services Endogenous variable AM AA AKS AUs ARS City growth attributes Percent urban 0.50 - 0.26 0.03 -0.01 - 0.03 City growth rate 10.29 -5.33 0.68 -0.26 - 0.67 In-migration rate 20.57 -10.65 1.36 -0.52 -1.33 Squatter house rents 3.57 -0.73 0.40 0 -0.14 Cost-of-living differential (urban + rural) 1.49 -0.80 0.12 -0.11 0.01 Selected economywide attributes PKS 0.82 0.22 -0.91 0.03 0 PUs 1.03 0.24 0.15 -1.09 0 PRS 0.02 1.17 0.09 -0.01 -1.03 QM 2.34 -0.57 0.24 0.08 -0.09 QKS 0.02 0.17 0.99 0.01 0 Qus 0.23 0 0.02 0.75 -0.01 QA -0.34 1.31 0.07 0 0.07 QRS -0.11 0.32 -0.01 0 0.60 LM 1.36 -0.84 0.10 0.18 -0.09 LKS -0.18 0.15 -0.07 0.11 0 Lus 0.24 0 0.02 - 0.26 -0.01 LA - 0.37 0.15 -0.02 0.01 0.08 LRS -0.11 0.34 -0.01 0 - 0.41 Source: Kelley and Williamson (1984b), table 4.1, p. 103. Note: Variables are defined as follows:AM, total factor productivity growth in the manufacturing sector; AA, total factor productivity growth in the agricultural sector; AKs, total factor productivity growth in the modern services sector; AUS, total factor productivity growth in the informal urban services sector;ARs, total factor productivity growth in the informal rural services sector; PKS, price per unit of modem services; Pus, price per unit of informal urban services; PRS, price per unit of informal rural services; QM, output in the manufacturing sector; QKS, output in the modern services sector; Qus, output in the informal urban services sector; QA, output in the agricultural sector; QRS, output in the informal rural services sector; LM, unskilled labor in the manufacturing sector; LKS, unskilled labor in the modem services sector; Lus, unskilled labor in the informal urban services sector; LA, unskilled labor in the agricultural sector; LRS, unskilled labor in the informal rural services sector. response is constrained, partly by problems with absorp- bottlenecks (the skilled wage rises) and capital scarcity tion of output, partly by short-run capacity constraints (the return to capital in manufacturing rises far above its and skill bottlenecks, and partly by a rising supply price return in other uses), which suggests that the long-run of unskilled labor in the cities. Table 3-9 focuses impact of unbalanced productivity advance on city on the rising supply price of urban labor. Urban job growth is far greater than these short-run elasticities creation fosters in-migration, but with some limitation. imply. Although land use shifts to squatter settlements to accommodate rising density, urban rents nevertheless rise steeply in the face of the migrant influx. Since most Scarcities of Fuel and Imported Raw Materials new in-migrants are unskilled and poor, rents in squat- Since urban-based manufacturing uses intensively ter settlements rise more sharply than do rents for fuels and raw materials (importables which we call Z- high-quality housing. Squatter-settlement rents reflect goods), any increase in the price of imported Z-goods excess demand for housing and sites; the rents for high- penalizes manufacturing directly and other urban activi- quality housing mostly reflect increased land scarcity as ties indirectly and thus inhibits urban job creation and the needs of the poor are partially accommodated by city growth. Even though the model admits the possibil- shifting land use. The cost-of-living advantage of the ity of direct and indirect substitution away from the rural area rises sharply as a result of increased urban more expensive imported fuels and raw materials, table rents. All these factors raise the average unskilled wage 3-10 reports that in-migration and city growth are still in the cities and tend to restrict city growth in the short constrained because urban activities tend to be Z- run. Furthermore, job creation is constrained by skill intensive. The elasticity of Pz, however, is low compared 40 Allen C. Kelley and Jeffrey C. Williamson Table 3-9. Short-Run Constraints on the City "unproductive" capital accumulation on employment Growth Response to Unbalanced Productivity demand, urban job creation, and city growth. Accumula- Advance (1970 Elasticities) tion in the urban modem sectors fosters job creation, Endogenous variable AM AKS and an investment policy which favors manufacturing at the expense of agriculture fosters urbanization. It is City growth attributes surprising, however, that accumulation of urban squat- City growth rate 10.29 0.68 ter housing (Hus) has the most potent short-run impact on urban job creation and city growth. This result Congestion indicators occurs even though the impact multipliers exclude the Percent urban land in squatter settlements 1.25 0.13 employment effects of the formation of the housing Squatter house rents 3.57 0.40 stocks through investment (and investment activities in High-quality house rents 1.16 0.15 Cost-of-living differential (rural s urban) 1.49 0.12 squatter dwelling construction are highly labor- intensive). The relatively large urban employment Factor market disequilibrium indicators effects associated with the accumulation of squatter Average urban unskilled wage 1.03 0.15 housing are therefore all indirect: housing rents de- Rural unskilled wage 0.02 0.09 Urban skilled wage 1.28 0.13 Return to capital in M 2.34 0.27 Return to capital in KS 0.80 0.07 Table 3-10. Factors That Influence City Growth Return to capital in A - 0.30 0.06 in Developing Countries (1970 Elasticities) Source: Kelley and Williamson (1984b), table 4.2, p. 105. Endogenous urbanization variable Percent City growth In-migration with the unbalanced productivity advance indicators Exogenous variable urban rate rate (the Ajs) or with the other two exogenous prices re- Land and labor ported in table 3-10. At first glance this may suggest that RA -0.03 - 0.69 -1.38 the sensitivity of urbanization to scarcity of fuels and Ru 0.04 0.88 1.77 raw materials has been overstated in the literature. In L -0.57 6.38 12.75 view of the historical record since 1960, however, one Accumulation must be cautious. After all, Pz soared in the 1970s after KM 0.09 1.82 3.63 its recorded stability during the 1960s. Thus, Pz may KKS 0.02 0.40 0.81 have been a dominant source of developing-country KA -0.03 -0.68 -1.35 urbanization over the past two decades in spite of the HKS 0 0.07 0.15 modest elasticities reported in table 3-10. (The issue is HuS 0.39 7.92 15.83 discussed again below.) HRS -0.13 -2.75 -5.50 Prices PZ -0.04 -0.89 -1.77 Price Policy and World Market Conditions PM 0.54 11.13 22.23 Table 3-10 presents the short-run impact multipliers PA - 0.32 -6.51 -13.01 for prices of agricultural and manufactured goods, both Productivity advance of which compete in world markets. These prices are AM 0.50 10.29 20.57 usually heavily distorted by external and internal price AA - 0.26 -5.33 - 10.65 policy. The table shows that city growth was far more AKS 0.03 0.68 1.36 sensitive to PA and PM than to Pz. Any effort to under- Aus - 0.01 - 0.26 - 0.52 stand the sources of past and future city growth must ARS - 0.03 - 0.67 -1.33 therefore carefully sort out these relative price condi- Other tions, including the impact of past liberalization and Skilled labor force (S) 0.23 6.96 13.90 future protectionist trends in the industrial countries Foreign capital (F) 0 - 0.03 - 0.06 and the influence of price-distorting policy regimes in Note: Variables, in addition to those defined in table 3-8, are as developing countries. follows: RA, agricultural land stock; Ru, urban land stock; L, total developing countries, unskilled labor force; KM, physical (productive) capital in the manufac- turing sector; KKS, physical capital in the modem services sector; KA, physical capital in the agricultural sector; HKs, high-quality urban Accumulation, Capacity, and Job Creation housing; Hus, low-quality (squatter) urban housing; HRs, rural housing; Pz, price per unit of imported raw materials; PM, price per Table 3-10 also summarizes the impacts of "produc- unit of manufactured goods; PA, price per unit of agricultural goods. tive" capital accumulation and population-sensitive Source: Kelley and Williamson (1984b), table 4.6, p. 111. What Drves City Growth in the Developing World? 41 crease with the augmented supply of dwelling space, the Table 3-11. The Growth Environment, Pre-OPEC relative cost of living in the cities declines, in-migration and Post-OPEC: Dynamic Parameters Assumed is fostered, nominal wages of the unskilled are lowered in the Sinulations by the temporary labor glut, and employment expands (percent) everywhere in the city as a result, especially in manufac- Average annual growth turing. In contrast to Coale and Hoover's (1958) empha- Exogenous variable Pre-OPEC Post-OPEC sis on the tension between unproductive and productive (dynamic parameter) (1960-73) (1973-79) capital accumulation in city growth, table 3-10 suggests no conflict: of the six alternative modes of accumulation Price of imported raw materials (Pz) 0 5.2 listed, accumulation in HUS has the highest urban job Price of M-goods (PM) -0.7 - 1.6 creation and city growth effects. An issue for future work Agricultural land stock (RA) 1.0 0.5 is to assess whether this conclusion holds for the longer Urban land stock (Ru) 1.0 1.0 run. Total labor force (L *S 2.54 2.68 Source: Kelley and Williamson (1984b), table 5.1, p. 126. Land and Labor Popular accounts of rapid urbanization in developing countries emphasize the role of high population growth. in the model by exogenous trends in PA, PM, andPz (table Table 3-10 contradicts this view, although how wrong 3-11). In relation to prices of primary exports from the population account is depends on which aspect of developing countries, the quality-adjusted price of urbanization and city growth is of interest. manufactured goods declined at an annual rate of 0.7 Lewis (1977) has stressed the capital intensity of percent up to 1973. The rate of decline accelerated after cities, and the present model conforms with that reality, 1973 and averaged 1.6 percent a year to 1979. In con- since urban activities are, on average, far less labor- trast, the price of imported raw materials (including intensive than are rural activities. The Rybczynski fuels) rose by 5.2 percent a year after 1973 compared theorem in trade theory holds that an increased endow- with PA; the same relative price exhibited long-run sta- ment of any given factor of production should favor the bility before 1973. Although these price trends are expansion of those sectors which use that factor most affected by the base periods selected, the averaging de- intensively. Thus, whatever its source, population- vices applied, and the underlying price series, there can induced labor force growth should foster the expansion be no doubt about the epochal character of the post-OPEc of rural activities and suppress urbanization. According price trends which developing countries faced in the to this analysis, population growth does not offer an 1970s. explanation for urbanization, and the negative impact As table 3-11 shows, the growth rate of the labor force multiplier in table 3-10 (percent urban, - 0.57) proves was higher after 1973, although the acceleration was the point. For in-imigration and city growth, however, quite modest. Table 3-11 also documents the assumed table 3-10 reports a more conventional result, since the rates of farmland growth, showing a decline which cap- impact multipliers are positive and quite large. (The tures the apparent rapid exhaustion of possibilities for next section elaborates on this issue.) augmenting extramarginal land. Clearly, there were What about land endowments? Conventional wisdom other (perhaps less important) nonprice changes in the has argued that scarcity of agricultural land has tended economic and demographic environment that sur- to push labor into the cities. Although an extension of rounded the RDE group after 1973. For that matter, not the arable land stock would certainly increase the reten- all of the epochal price trends were related to OPEC. tion of labor in rural areas and thus retard urbanization, Nonetheless, we have labeled these two epochs pre-oPEc the size of the impact reported in table 3-10 is small. and post-oPEc. Changes in agricultural land endowment are simply not an important part of the city growth tale. Evaluatng the OPEC Watershed The world economy has been undergoing painful The OPEC Watershed and Recent adjustments to the price shocks associated with short- Growth Trends run OPEC policy and long-run scarcities of raw materials and fuels. Since as late as 1979 the world economy was The changing growth environment in the aftermath still digesting the impact of these disequilibrating of the OPEC price shock has been significant. It has been shocks, it seems clear that a long-run general equilib- manifested mainly in relative prices, which are captured rium model such as ours cannot be expected to account 42 Allen C. Kelley and Jeffrey G. Williamson adequately for the short-run trends that developing tions by the model in table 3-13 might suggest that the countries have undergone since 1979. Nonetheless we OPEC price shocks mattered little to subsequent urban can use the post-oPEc conditions to illustrate the mod- performance in the late 1970s. The annual rate of city el's sensitivity to such epochal shocks. growth declines only modestly over the period as a whole Tables 3-12 and 3-13 summarize the city growth pre- (from 4.86 to 4.65 percent); the percent urban continues dictions. Table 3-13 offers a 1960-73 prediction, using its rapid climb; and in-migration rates, although some- the actual pre-OPEC economic and demographic environ- what lower, remain high. Note, however, the unambig- ment, that has already been compared with actual ex- uous evidence of retardation in the Actual columns of perience in previous sections; a 1973-79 prediction us- table 3-12, in which the rates of in-migration and city ing the actual post-OPEc economic and demographic growth lose their steam quite dramatically. More impor- environment documented in table 3-11; and a counter- tant, in table 3-13 compare the predicted actual post- factual post-OPEC prediction, which simply allows the OPEc experience with the counterfactual 1973-79 experi- pre-OPEC environmental conditions to continue beyond ence that assumes pre-OPEC environmental conditions: 1973. The counterfactual experiment makes it possible while the actual city growth rate for 1973-79 is 4.65 to assess what urbanization would have been like with- percent a year, the counterfactual rate would have been out the post-OPEC epochal shocks. 6.04 percent. Furthermore, without the shocks of the Comparison of the pre-1973 and post-1973 predic- late 1970s, the counterfactual rate of in-migration Table 3-12. Post-OPEC Urban Adjustnents, Model Predictions, 1973-79 Actual post-OPEC environment Counterfactual environment Net urban Net urban Percent City in-migration Percent City in-migration Year urban growth rate urban growth rate 1973 43.45 5.72 3.07 43.45 5.72 3.07 1974 44.45 5.10 2.32 44.78 5.75 3.10 1975 45.22 4.48 1.72 46.25 5.92 3.27 1976 46.26 5.03 2.26 47.80 6.03 3.38 1977 47.05 4.52 1.76 49.46 6.14 3.47 1978 47.85 4.47 1.71 51.22 6.23 3.56 1979 48.58 4.29 1.53 53.03 6.16 3.49 Note: The counterfactual results assume that the 1960-73 dynamic parameters persist after 1973. The net urban in-migration rate is the ratio of annual in-migration to the average urban population in all previous years. Source: Kelley and Williamson (1984b), table 5.3, pp. 130-31. Table 3-13. Pre-OPEC and Post-OPEC Urban Adjustments, Model Predictions, Period Averages (percent) Basis for prediction Actual Actual pre-OPEC post-OPEC Counterfactual environment, environment, environment, Variable 1960-73 1973-79 1973-79 Annual city growth 4.86 4.65 6.04 Total increase in share urban for period 10.85 5.13 9.58 Annual increase in share urban 0.83 0.86 1.60 Net in-migration share of urban population increase 47.72 41.34 57.50 Net in-migration rate 2.35 1.83 3.30 Net out-migration rate 1.37 1.60 3.13 Note: The 1973-79 counterfactual results assume that the 1960-73 dynamic parameters persist after 1973. The net in-migration rate is the ratio of annual in-migration to the average urban population in all previous years. The net out-migration rate is the ratio of annual rural out-migration to the average rural population in all previous years. Source: Kelley and Williamson (1984b), table 5.3, pp. 130-31. What Drives City Crowth in the Developing World? 43 would have risen to 3.30 percent rather than fall to 1.83 certainly has contributed to spectacular city growth percent, as in fact happened. rates, but the experiments suggest that it is not the central force that drives urbanization. Table 3-14 also suggests that a shift to foreign capital austerity would in City Growth not matter much to urbanization. Urbanization in developing countries is, however, It seems clear that exogenous economic and demo- sensitive to a productivity slowdown. Given a plausible graphic conditions had a powerful impact on develop- retardation in the economywide rate of TFPG from 1.8 to ing-country urbanization during the 1970s. Table 3-14 1 percent a year, city growth rates decline significantly, reports an effort to isolate the most important factors. and the impact seems to show signs of cumulating over Nine counterfactual stimulations are used. Each gener- time. ates a history in the 1970s (1973-79), but the table presents only one of the model's predictions-the annual rate of city growth. Each counterfactual case Conclusion should be compared with the actual 1973-79 perfor- mance reproduced in column 1 (which repeats table This chapter has shown the ability of the Kelley-Wil- 3-12). The fuel abundance counterfactual in column 3 liamson model to adequately replicate growth, accu- maintains all of the exogenous conditions that underlie mulation, distribution, and city growth in developing the actual performance in column 1 except*fuel price countries up to the 1973 watershed. On that basis it is behavior: while the oPEc-augmented actual Pz was 5.2 concluded that the model can be used to analyze the percent a year between 1973 and 1979, the counterfac- sources of urban growth during the 1960s, the early tual assumes that Pz = 0, as was indeed the case up to 1970s, and the difficult period of structural adjustment 1973. since then and to analyze future trends in urban experi- The urban slowdown had little to do with agricultural ence under varying world market conditions and domes- land expansion or labor force growth. Rather, it appears tic policy regimes. that prices were doing most of the work. Furthermore, To summarize, rapid rates of population growth are fuel scarcity, in spite of the attention it has received, was not-as a reading of the popular literature would sug- not nearly as important a source of slowdown in city gest-the central influence that drives rapid urban growth as was the accelerated decline in the relative growth in developing countries. Capital transfers to de- price of manufactures (lower PM) This finding obviously veloping countries have not played a significant part, supports the view that any future trend toward protec- and rural land scarcity has had only a modest role. The tionism in industrial countries will play an important most potent influences on city growth appear to have role in shaping city growth in developing countries in been the rate and imbalance of sectoral productivity the next two decades. The same might be said, of course, advances-technological events which have favored the of the mix of internal policies which may twist the urban modern sectors-and prices (although oPEc- relative price of manufactures in the future. This posi- induced fuel scarcity has been less important than the tion is reinforced by the counterfactual experiment relative price of manufactures). Thus, trade policy in the under "stable world markets," in which the relative industrial countries and price policy in developing coun- price of manufactures is held fixed during the 1970s. tries are likely to have the most important impacts on Under these more favorable price conditions for manu- city growth in the next two decades. facturing the model predicts annual city growth rates of 6.49 percent, in contrast to the actual rate of only 4.65 percent. Although a sharp decline in the rate of population Notes growth would certainly have diminished the rate of city growth, column 8 in table 3-14 suggests that this in- 1. Algeria, Bangladesh, Brazil, Cameroon, Chile, Colombia, fluence has been grossly overdrawn in the popular litera- Costa Rica, Cote d'lvoire, Dominican Republic, Ecuador, Arab ture. The counterfactual explores the impact of a spec- Republic of Egypt, El Salvador, Ethiopia, The Gambia, Guate- tacular reduction in population pressure from the actual mala, Honduras, India, Indonesia, Kenya, Korea, Malaysia, RDcEa raeduction8 iecnt popuartionrsuefo the a. e ctualic Mexico, Morocco, Nicaragua, Nigeria, Pakistan, Panama, Para- RDE rate of 2.68 percent a year to the 0.9 percent which guay, Peru, Philippines, Portugal, Sri Lanka, Swaziland, Syria, prevailed among industrial countries. Even under this Taiwan, Thailand, Togo, Turkey, Uganda, and Yugoslavia. enormous diminution in population growth, the rate of 2. The World Bank's World Tables normally report only city growth would still have reached almost 4 percent a total private savings and do not separate household savings year for the period 1973-79. Rapid population growth from firm (corporate) reinvestment. Although the distinction Table 3-14. Sources of a Slowdown in City Growth, 1973-79 (percent) Other counterfactuals, 1973-79 OPEC watershed counterfactuals, 1973-79 Technological Population No population slowdown pressure, pressure (decrease in Foreign Total Fel abundance, World markets, Land expansion, pre-OPEC, Stable (developed TFPG from capital pre-OPEC ~~~~~~~~~~pe-PE,pre-OPEC, pr*e-OPEC, Actual peOE pr-PCppre-OPEC, reOE,*world markets, country l.S to I austerity, 1973-79 environment Pz = 0 only PM only RA only L + S only PM= 0 rate) percent) F= 0 Itemn (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Year 1973 5.72 5.72 5.72 5.72 5.72 5.72 5.72 5.72 5.72 5.72 1974 5.10 5.75 5.35 5.59 5.09 5.06 5.95 4.46 5.10 5.21 1975 4.48 5.92 4.91 5.67 4.51 4.50 6.46 4.03 4.27 5.66 1976 5.03 6.03 5.28 5.90 4.95 4.96 6.51 4.37 4.60 4.37 1977 4.52 6.14 5.13 5.91 4.47 4.48 6.64 3.72 4.22 4.33 1978 4.47 6.23 5.05 5.96 4.36 4.36 6.63 3.68 3.93 4.21 1979 4.29 6.16 4.83 5.79 4.27 4.28 5.76 3.57 3.82 4.24 Average 4.65 6.04 5.09 5.80 4.61 4.60 6.49 3.97 4.32 4.67 Exogenous variables taken to be: PZ 5.2 0 0 5.2 5.2 5.2 5.2 5.2 5.2 5.2 PM -1.6 -0.7 -1.6 -0.7 -1.6 -1.6 0 -1.6 -1.6 -1.6 RA 0.5 1.0 0.5 0.5 1.0 0.5 0.5 0.5 0.5 0.5 L + 5 2.48 2.54 2.68 2.68 2.68 2.54 2.68 0.9 2.68 2.68 L+S F such that F/GDP= 3 percent TFPC 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.8 1.0 1.8 Note: Variables are defined as follows: Pz, fuel and raw materials prices; AM, domestic price of manufactures; RA, agricultural land stock; L population (or labor force); F, foreign capital inflow; TFPc, economywide productivity growth. Sources: Kelley and Williamson (1984a), table 5, p. 435; (1984b), table 5.4, p. 133. What Drives City Growth in the Developing World? 45 may be fuzzy in the economies being analyzed, the two sources 1984a. "Population Growth, Industrial Revolutions, of savings are quite distinct in the model and are thus reported and the Urban Transition." Population and Development separately. Furthermore, what the model identifies as training Review, vol. 10, no. 3 (September), pp. 419-44. investment is embedded in private and government consump- . 1984b. What Drives Third World City Growth? A tion in the RDE national accounts. General Equilibrium Approach. Princeton. N.J.: Princeton University Press. Bibliography Keyfitz, Nathan. 1980. "Do Cities Grow by Natural Increase or by Migration?" GeographicalAnalysis, vol. 12, no. 2 (April), Becker, Charles M., Edwin S. Mills, and Jeffrey C. Williamson. pp. 142-56. 1986. "Modeling Indian Migration and City Growth, 1960- Ledent, Jacques. 1980. Comparative Dynamics of Three De- 2000." Economic Development and Cultural Change vol. mographic Models of Urbanization. RR-80-1. Laxenburg, 35, no. 1 (October), pp. 1-33. Austria: International Institute for Applied Systems Analy- Beier, George, Anthony Churchill, Michael Cohen, and Ber- sis. February. trand M. Renaud. 1975. The Task Ahead for the Cities of Lewis, W. Arthur. 1977. The Evolution of the International Developing Countries. World Bank Staff Working Paper 209. Economic Order. Princeton University, Woodrow Wilson Washington, D.C. Also in World Development, vol. 4, no. 5 School, Research Program in Development Studies, Discus- (May 1976), pp. 363-409. sion Paper 74. Princeton, N.J. Chenery, Hollis, and Moshe Syrquin. 1975. Patterns of De- Linn, Johannes F. 1979. Policies for Efficient and Equitable velopment, 1950-1970. London: Oxford University Press. Growth of Cities in Developing Countries. World Bank Staff Coale, A. J., and E. M. Hoover. 1958. Population Growth and Working Paper 342. Washington, D.C. Economic Development in Low-Income Countries. Prince- Mohan, Rakesh. 1979. Urban Economic and PlanningModels: ton, N.J.: Princeton University Press. Assessing the Potential for Cities in Developing Countries. Kelley, Allen C., and Jeffrey G. Williamson. 1980. Modeling Baltimore, Md.: Johns Hopkins University Press. Urbanization andEconomic Growth. RR-80-22. Laxenburg, Preston, Samuel. 1979. Patterns of Urban and Rural Popula- Austria: International Institute for Applied Systems Analy- tion Growth. New York: United Nations, Population Divi- sis. May. sion. . 1982. "The Limits to Urban Growth: Suggestions for United Nations. 1976. GlobalReview ofHuman Settlements:A Macromodelling Third World Economies." Economic De- Support for Habitat. Vols. 1 and 2. Oxford: Pergamon Press. velopment and Cultural Change, vol. 30, no. 3 (April), pp. World Bank. 1976. World Tables 1976. Baltimore, Md.: Johns 595-623. Hopkins University Press. Part II A Framework for Urbanization Policy The explanations of urbanization suggested in part I externalities such as pollution and congestion cause provide a framework of economic development cities to be seen as "too big." Measurement of these within which to discuss problems connected with urban- effects, however, reveals that the net economic costs of ization. The policy issues addressed in part 11 are pre- overurbanization (compared with the altematives) are sented in the context of overall development and the surprisingly small. Furthermore, these externalities are macroeconomic settings of countries. Tolley (chapter 4) best dealt with directly rather than by trying to control goes to the economic sources of widely noted urban urbanization or city size. A more satisfactory size dis- problems, Renaud (chapter 5) discusses policy experi- tribution might also come about in the absence of urban ences from broad countrywide and cross-country per- biases and such incentives as protected employment and spectives. This broad approach is intended not to special subsidies in urban areas. minimize the urban-specific sources of various urban In developing countries the disproportionate concen- problems, but to bring out some important and some- tration of populations in big cities makes the scale of times neglected overall considerations which have per- overurbanization more dramatic than it has ever been. vasive effects on the urban economy. Parish (chapter 6) One reason for urban concentration is that transport examines urban problem-solving in China's centralized costs within cities are lower today than when the devel- policy framework. oped countries were undergoing rapid urbanization. Be- cause economies of association can now be realized more cheaply within a city, it is easier to locate ancillary Urban Problems and Policy Implications local production and the needed housing in the city. Infrastructure planning and construction may also con- The most basic urban problem, as noted in chapter 4, tribute to disproportionate expansion in cities that are is poverty. The extent of urban poverty, in the first already large, as growth feeds on growth. instance, can be understood in the context of the rate and nature of economic development and the generation of employment. Given the level of development, rural- urban migration helps to reduce differences in real in- The Scope of Urbanization Policy comes between urban and rural areas and bring about an increase in urban poverty. Sustained and rapid develop- Chapter 5 provides an account of the range of analyt- ment is in most instances a necessary, although not ical and policy problems that arise in the context of always a sufficient, condition for alleviating urban pov- developing-country urbanization, with examples drawn erty. Attention to the distribution of the benefits of from different continents. The author also examines the development is also essential. differences between the urban policy problems that con- This is not to deny that urbanization can itself con- front the developing countries and those that the de- tribute to problems, including poverty. Many countries veloped countries face. Two structural changes in the face excessive urbanization and the associated problems developed world help explain the differences: the slow- of overcrowding and deplorable living conditions. In ing and eventual end of urbanization, and the emer- economic terms overurbanization occurs when negative gence of yet another industrial revolution, this one 47 48 Part 11 based on new technologies and less tied to concentrated The distinction between urban work in China and manufacturing centers. elsewhere has been highlighted by differences in the Problems in developing countries that are associated distribution of nonagricultural labor: the Chinese have with urbanization include spatial inequalities, such dis- emphasized work in manufacturing, mining, and con- amenities as congestion and pollution, and the need to struction and have downplayed consumption-related create employment and provide services. Urbanization activities. The consequences have been less varied and policies actually followed by developing countries can be less plentiful consumer goods and rationing and classified as national economic policies, explicit spatial queuing. An additional problem has been the unemploy- and regional policies, and city management. Their rela- ment, especially among young people, that has accom- tive importance varies widely from country to country. panied investment in capital-intensive heavy industry. Problems of congestion, pollution, and internal city Many educated young workers were sent to the country- inefficiency are best addressed directly through good side to relieve urban unemployment, and unemploy- city management. The attempt to control city popula- ment rose to 5-11 percent of the nonagricultural labor tion directly is an inefficient and costly way of dealing force. Reforms after 1978 loosened investment policies with those problems. At the same time it should be and restrictions on work opportunities and made public recognized that national economic and sectoral policies education more responsive to the needs of industry. By often have spatial biases. Reversal of such biases, partic- the end of 1981, outright unemployment had fallen to ularly those that hurt rural production, will minimize 2.4 percent of the nonagricultural labor force. pressures to control city population directly. Explicit Another aim has been to provide adequate urban ser- spatial policies may seem desirable, particularly in situa- vices. The authorities have displayed ambivalence tions in which the nation's economic policies have, toward urban services and have agreed only that there intentionally or unintentionally, produced urban biases. should be an equitable distribution of basic needs. The efficacy of direct spatial initiatives must be reviewed Medical services have grown steadily and are of higher carefully, however, particularly since experience with quality than in other developing countries; infant mor- them is not encouraging. tality is lower and life expectancy higher. Urban housing has received less attention, partly because it has been A Case of Central Planning considered a consumption good, but Chinese urban housing compares favorably with that in other develop- Chapter 6 uses data from a variety of sources to piece ing countries. The study discusses the unintended side together a view of modern cities in China and sheds light effects of rent control and the recent sharp increases in on some major goals of Chinese urban policy. One goal state funds to remedy shortcomings in housing. has been balanced city size distribution and stability. In the 1950s investment funds and activities were chan- Bibliography neled from coastal cities and reinvested in the interior. Tough migration laws and near-total control of social Birdsall, Nancy. 1980. Population and Poverty in the Develop- institutions by local and central authorities furnished ing World. World Bank Staff Working Paper 404. Washing- the necessary muscle. The study offers a detailed review ton, D.C. of recent urbanization in China, which contrasts with Henderson, J. Vernon. 1977. Economic Theory and the Cities. that in most developing countries. New York: Academic Press. Another goal has been secure, productive urban em- Lipton, Michael. 1977. Why Poor People Stay Poor: Urban Bias ployment, guaranteed by the state and devoted to cen- and World Development. Cambridge, Mass.: Harvard Uni- trally defined output objectives. Several reasons are versity Press. cited for the rise in the employment rate of the nonagri- Mills, Edwin, and Charles Becker. 1982. "Urbanization and cultural population: the eviction from cities of unautho- Economic Development." Princeton, N.J.: Princeton Uni- rized people, the increasing employment of women, the versity. Processed. sharp drop in urban birth rates, and efforts to create new Thomas, Vinod. 1982. Differences in Income, Nutrition and urban employment opportunities. The largest em- Poverty within Brazil. World Bank Staff Working Paper 505. ployers of the urban population are the various levels of Washington, D.C. government, and state employees in China seem to en- World Bank. 1979. World Development Report 1979. New joy greater benefits than do workers in other developing York: Oxford University Press. countries. In the smaller employment sectors, collective . 1981. World Development Report 1981. New York: enterprises and the private work force, wages are lower, Oxford University Press. benefits are fewer, and employment is less secure than in . 1982. World Development Report 1982. New York: the state sector. Oxford University Press. 4 Market Failures as Bases of Urban Policies George S. Tolley U rbanization in developing countries remains con- ing direct spatial policies to influence urbanization are U troversial. At the heart of many economic develop- discussed, and the role of externalities in nonspatial ment efforts has been promotion of industrialization, policies is considered. The need to improve methods for which fosters the growth of urban centers. But urban- evaluating the spatial effects of infrastructure decisions ization has also been viewed as undesirable because it is emphasized, and evaluation needs in the management takes people away from a rural way of life that is thought of urban services in general are brought out. to be intrinsically better and because of the economic and social problems found in cities. Urban poverty has been viewed as one of the ills of Urbanization and Poverty growth. Recurrent assertions that overurbanization has taken place refer to environmental problems in cities, The "rings of thorns"-or shantytowns on the out- wage policies that induce excess rural-to-urban migra- skirts of many of the large cities of the developing tion, and government and tax policies that favor cities. world-have been the subject of much concern, and The arguments have usually been made in qualitative they are, at least subconsciously, partly responsible for terms and have seldom been related seriously to the the view expressed by some that excessive urbanization functioning of markets. has occurred. The growth of the numbers of poor people This chapter offers an analysis of these and other is an especially noted symptom of urbanization, and issues that bear on urbanization performance. The there may be a lurking hope that curbing urbanization causes of urban poverty are first brought out, and the would reduce poverty. relation of poverty to overall economic development Interest persists in estimating the numbers of people considerations is discussed. Attention then turns to in poverty according to some definition. Aside from the accounting problems that distort measured changes in problem of choosing the level of income below which national income when urbanization occurs. poverty is deemed to exist (a workable procedure if the Overurbanization is then analyzed. Models with choice is recognized as arbitrary), the problem arises of measurement are developed that deal with pollution, measuring the extent of poverty among different groups congestion, urban wages, and government fiscal poli- within the population. That task is made difficult by the cies. The models give indications of the city size effects varying circumstances of people, which make simple and income losses attributable to urbanization external- comparisons of money income invalid. ities. The reasons why large cities in developing coun- Explaining the extent of urban poverty requires an tries are so large and are getting larger are analyzed, understanding of what people earn and is more difficult with special attention to transport. than explaining the overall degree of urbanization (dis- The final section draws policy implications. Possibili- cussed in chapter 2). Yet a foremost reason why poor ties are considered for internalizing externalities as a people are found in large numbers in cities of developing first-best policy that is superior to trying to influence countries is the sheer overall magnitude of poverty in urbanization as such. Options and pitfalls in undertak- those countries and its distribution among urban and 49 50 George S. Tolley rural areas because of migration. Through migration, farming, the demand in urban areas consists of demands rough equalization of real income levels between urban for common labor in a variety of activities, including and nonurban areas can be expected to be achieved factory work and services. Services in urban areas are within each income stratum, from rich to poor. Lack of supplied through work in formal firms and as domestic knowledge of opportunities in different areas is not an help as well as in casual and informal employment. apparent barrier to migration, nor is lack of such con- Common labor is itself an aggregation and simplifica- tacts as friends and relatives who can help the migrant tion that masks many gradations of productivity within get settled in a new environment. the unskilled category. In speaking of the poor, one is Equalization of real earnings between rural and urban speaking primarily of familieswhose breadwinners are at areas may not be exact. Whether equalization takes place the lower ends of the unskilled distribution. cannot be proved or disproved by looking at urban and Although, as noted, the demand for unskilled labor is rural money incomes at any one time. Observed money affected by accumulations of physical capital and growth incomes are not directly commensurate in rural and in in the number of skilled laborers, these two types of urban areas. People place different values on differences productive factors will also tend to grow as the output of in the availability of goods, prices, and general environ- urban commodities rises. The number of unskilled ments. Thus the equalization is not between money laborers, the number of skilled laborers, and the amount incomes but rather between real incomes, including the of physical capital will all tend to grow with urbaniza- values people place on differences between rural and tion. Performance levels of skilled laborers may rise, urban areas. owing to education and to experience with more sophis- The hypothesis that a rough equalization of real earn- ticated techniques, and physical capital may accumulate ings takes place is supported by two arguments. The first more rapidly than population. These developments may is introspective: it does not stand to reason that people, affect the demand for unskilled labor, but they operate to in making choices about where to live, will ignore some extent in rural as well as in urban areas. opportunities in other places. If they pay some attention Changes in the numbers of unskilled laborers in to alternatives, differences in real earnings between urban and rural areas often go along, more or less, with places will not get far out of line. The second argument is changes in urban and rural populations. Without claim- empirical: if people did in fact ignore real income differ- ing an exact proportionality, it seems that changes in ences, the erratic pace of demand for products and unskilled labor and population in an urban area tend to supplies by people in different places, as determined move similarly. If so, the explanation of quantitative by demographic considerations, would produce wide changes in percentage terms for the urbanization of swings in observed relative differences in money in- unskilled labor is similar to that already considered for comes. Although some annual fluctuations in the ratio the population as a whole. In some countries lower- of money incomes between rural and urban areas take income and subsistence agriculture become outmoded, place, over the years the fluctuations are limited. Thus, leading to a low or even negative apparent income elas- explaining poverty in cities becomes a matter of explain- ticity for unskilled labor in agriculture. In this case the ing how the demand for the services of poor people in general approach to the explanation of urbanization of urban areas interacts with their conditions of supply. the unskilled population remains the same as for the The same type of explanation applies to people in rural total population, but the quantitative outcome is differ- areas. The supplies in each case are determined by the ent. The implication in this case is that the rate of total number of persons of a given type in the economy. urbanization of the unskilled population will be more Suppose we letNu and Na refer not to the total number rapid than that of the population as a whole. of persons in urban and rural areas, as in chapter 2, but There are three reasons why the absolute level of to numbers of persons who supply unskilled or common poverty diminishes during development. First, the ap- labor in the two types of area. Then the same kind of plication of new techniques and capital raises the mar- explanation of the rate of urbanization of the total ginal productivity of unskilled labor. This phenomenon population that was given in chapter 2 can be used to is in many ways the essence of development. If develop- explain the rate of urbanization of unskilled labor. The ment does not succeed in raising the well-being of all income elasticities will reflect the income elasticities of strata, it does not succeed in its purpose. Moreover, the final products in which the unskilled labor is used, looking at the various gradations of development around and they will also reflect the effects of the accumulation the world, one sees that a primary distinguishing feature of physical capital and the growth in numbers of skilled among countries is the extent to which the level of laborers on the demand for unskilled labor. Whereas the well-being of all strata appears to be raised. demand for unskilled labor in rural areas consists pri- It may be hypothesized that absolute poverty tends to marily of demands for field work and lower-income decline at first slowly and then at an accelerating rate. If Market Failures as Bases of Urban Problems 51 labor is paid its marginal product, the percentage rate of disappears. All that is left on the right-hand side is the increase in wages of unskilled labor-which determines weighted sum of the increases in marginal productivity the rate of decline of poverty-will equal the percentage that would occur in the absence of changes in amounts rate of increase in marginal productivity. The percent- of labor in the different activities. The percentage age rate of increase in marginal productivity depends in change in wage is turn on changes in techniques and in quantities of (41) 1 = nuA +n,A0. associated factors of production that directly raise mar- ginal productivity and on changes in the proportions of Refinements to allow for differences in elasticities of labor and other factors of production used in a given demand for labor could be introduced, but the terms activity. If more unskilled labor is used with the other shown would probably still dominate the results. The factors of production, its wage will go down, and if less is result shown generalizes to many activities and can be used itswagewill go up. Forthe economy as awhole, any written wj = ln1A,. increase in unskilled labor in one activity must be The percentage change in the marginal product of accompanied by decreases elsewhere, assuming that labor in agriculture, A0, may be small, either because there is a given amount of unskilled labor in the econ- progress in agriculture is slow or because a low-income omy at any one time. Moreover, because real incomes for or subsistence sector makes up a large part of agricul- unskilled labor tend to be equalized among activities, ture and is not progressing. At the early stages of de- the percentage rate of change in wages of unskilled labor velopment, when most labor is in agriculture, n0 will be will tend to be the same in different activities. close to one and nu will be close to zero. Even though Combination of these conditions gives insights on there is rapid progress in urban productivity, and hence how wages of unskilled labor in an economy change over a large value of A,,, the percentage rise in the wage will time. The percentage change in the wage, tw, will equal be small because of the low weight of A,,. As urbaniza- the increase in marginal productivity A (attributable to tion proceeds during development, the fall in n0 and the improvements in techniques and changes in amounts of rise in n,, will increase the weight ofA a and decrease the associated factors) plus the product of the elasticity of weight of Aa. As a result, the rise in the wage rate will demand for labor in an activity, ,B, and the percentage accelerate until, if the extreme of complete urbanization change in employment of labor in that activity, h. The occurred, the rise in the wage would be fully equal to the economy is then divided into urban and nonurban activ- upward shift in urban productivity. ities, and the conditions for the percentage change in Besides the rise in the return to unskilled labor, owing wage arew = Au + f3unu and& = Aa + ana.The first to improved techniques and the use of associated factors condition is multiplied by the proportion of the labor of production, two additional reasons for a decline in which is urban, nu, and the second condition by the absolute poverty may be noted. First, people acquire proportion which is nonurban, n0, and the results are skills and move out of the unskilled labor category. They added. Since the total of the proportions is unity (nfu + do this because economic development, by raising the na = 1), the left-hand side of each equation, after marginal productivity of skilled labor, increases the re- adding, equals iw. On the right-hand side, the elasticities turns to education and other human capital investments of demand for labor, 0,u and 03,, will be equal if the factor by which people transform themselves from unskilled to shares of the labor are the same in urban and nonurban skilled laborers. All persons who do so lift themselves activities, if their elasticities of substitution with other from poverty categories to a higher-income status. factors are the same, and if the price elasticities of The second reason is closely related to the first. After demand for the products produced are the same. These some people have transformed themselves into skilled assumptions are not likely to be fulfilled exactly, but laborers, there are fewer unskilled laborers in the econ- they serve as a useful benchmark approximation. The omy. Unskilled labor becomes scarcer than if the same assumptions will in fact be fulfilled if the product prices proportion of the population had remained unskilled. In are given, as may happen if the economy is importantly view of the elasticity of demand for labor, the greater involved in intemational trade and if the aggregate pro- scarcity of unskilled labor will lead to a rise in its wage. duction functions for the two types of activities are similar as regards shares and elasticities of substitu- tion-which again may not be unreasonable. If the elas- Urbanization as a Source of Growth: ticities are equal, one part of the right-hand side will An Illusion consist of two terms which combine to form the expres- sion nu,u + nnh0. But since h = dn / n and dn, + dn2 The fact that per capita money incomes are generally = 0, in view of the given amount of the labor in the higher in urban than in rural areas leads to an increase economy, nfuhu + n0it0 equals zero and the expression in measured per capita income under urbanization. In 52 George S. Tolley growth source analysis, which relates changes in na- foreseen. If the shifts are foreseen-and the fact that the tional output to changes in inputs, the rural-urban shift rural-urban shifts are large and of long duration during shows up as a contribution to growth because differ- development gives reason to believe that they are- ences in factor earnings are assumed to reflect real people can still make migration choices that equate real differences. The result is sometimes taken as an indica- earnings. A differential might be required to compensate tion that urbanization is itself a source of growth, and for the costs of moving, but empirically these costs thus that urbanization leads to additional development appear to be small. A factor that could actually make beyond that attributable to the demand and supply shifts urban wages lower in real terms than earnings in rural that have been analyzed. areas is the prospect of greater opportunity for increases That the rural-urban shift is a source of growth may in earnings in cities. If there is little opportunity for be partly or even wholly an illusion. A prerequisite of career advancement in rural areas, lifetime rural income meaningful interpretation of the rural-urban shift is that profiles will be flatter. The steeper profiles in urban areas education and other sources of differences in the average could lead migrants from rural areas to accept lower quality of labor in rural and urban areas be adequately entry-level earnings in urban areas. measured. Two concomitant changes in economic de- A final possibility, and perhaps the only actual source velopment are, ordinarily, a shift of less-educated labor of real effects of a rural-urban shift, is the delayed de- from rural to urban areas and an increase in education cline of subsistence-type agriculture. As emphasized levels. If the increase in education levels is greater in earlier, whole subsectors may become outmoded, in urban than in rural employment, the shift from rural to which case there is likely to be a generational process of urban employment will appear to give rise to growth replacement of human capital. Assuming that this large- unless the differences in education and other reasons for scale change was not foreseen, older persons may be skill differences are carefully controlled. What is really a found in low-income career paths in subsistence-type contribution to growth by increases in human capital agriculture even though their earnings are lower than may appear as a contribution by the rural-urban shift. earnings available elsewhere. Their offspring who re- Even after differences in quality of inputs have been place them in the labor force then migrate to cities and taken into account, measured contributions of the rural- find more productive employment. This intergenera- urban shift may remain which are illusory. The fact that tional shift from low-productivity to high-productivity nonurban areas have more home production and other employment leads to an increase in total national in- production which does not pass through the market- come. It should be noted that this is not an independent place and therefore cannot be measured is often pointed source of growth but rather a manifestation of a lag in out. adjustment; it reflects a failure to respond promptly to A perhaps more important consideration is that the past sources of growth. prices of local goods tend to be higher in urban than in nonurban areas. A prime example is residential rent, but the phenomenon also pervades services. Higher money Has Overurbanization Occurred? wages that make labor as well off in urban as in rural areas-as will tend to be brought about by migration- So far in this chapter no reasons have been found to will have a multiplier effect and will raise prices of local question market performance in response to the demand goods produced by the labor. The higher money wages and supply changes that determine urbanization. This have a feedback effect and raise money wages still more. section considers possible benefits and costs of the pro- Thus, in a market economy with labor mobility, the cess of urbanization that people's responses do not take prices of local goods can be expected to vary, and differ- into account. ences in money wages will appear even if real wages are equated in different locations. Differences in prices be- tween areas may be wrongly interpreted as causing changes in total output when there is a rural to urban As noted above, differences in living conditions be- population shift. tween rural and urban areas cause money income differ- Whether real effects of the rural-urban shift remain ences because of people's responses to these living con- after measurement is carried out correctly is moot. It is ditions. The question now is whether people, in their sometimes said that a differential in real eamings be- migration responses that influence the degree of urban- tween rural and urban areas is required to induce the ization, impose costs or gains on other people which shifts in resources that occur during growth. Expecta- those making the decisions have little or no incentive to tional considerations should be recognized, however. A consider. Two of the many costs which are not fully differential may be required only if the shifts are not intemalized are pollution and congestion. Market Failures as Bases of Urban Problems 53 Commuting costs, pollution, and congestion are impairment of aesthetic values. This damage value among the many reasons that money wages are gener- would imply that the cost of pollution to a family in the ally observed to be higher in urban than in nonurban larger city is $50 more (50 micrograms difference multi- areas. New entrants to a city pay the full costs of longer plied by $1) than the cost to a counterpart family in a commuting distances in the form either of long travel to smaller city with better air. If air pollution were the only work or of high rents if they do not commute from the unusually adverse characteristic of larger cities, and if agricultural edges of the city. New entrants likewise bear the population were stable, the wage in the city of higher costs of pollution and congestion than in rural 600,000 would have to be $50 more than in smaller areas, but only the average of such costs-not the extra places. costs that they impose on all others in the city because If the pollution effect is added to the distance effect, their presence leads to additional air pollution and slow- the estimated amount by which wages in the larger city ing of traffic. must exceed wages in smaller places to compensate for A way of approaching the question of market perfor- the disamenities of the larger city is $125. This is a mance is to ask how much higher money wages must be conservative estimate because many of the numerical in a larger city than in a smaller one to compensate for estimates used in the derivation are conservative and the increased costs in the former. As a simplified nu- because a number of disamenities (such as other pollut- merical example, take a developing-country city with a ants and noise, visual blight connected with slums, and population of 600,000 and a work force of 150,000 per- crime) have been neglected. If these figures were correct sons employed full-time at an average annual wage of and if no other factors influenced decisions to migrate, $2,000. people would disperse themselves among cities in such a Consider first the commuting distance effect. When a way that the wage rate was, say, $2,000 in smaller cities laborer migrates from farm to city, the edges of the city and $2,125 in the larger city to equalize the advantages are moved farther out into the agricultural periphery. of living in one or the other city. Anyone living in from the periphery must pay higher The factors considered so far represent average costs rents because of the commuting advantage of interior per worker or per family, which are higher in larger than residential sites over peripheral locations. As the city in smaller cities. The question becomes whether the grows money wages tend to increase in relation to addition of a worker in a city adds the same amount to money wages in nongrowing places. Suppose that in costs, or a greater amount. such a city each worker commuting from the periphery One reason the costs may be greater is that travel for of the city spent, on average, 15 minutes longer a trip everyone else in the city may be slowed by the extra traveling to work than he would from peripheries of congestion. Suppose that in a city of 600,000 half of the typical smaller cities. If the extra time is multiplied by labor force, or 75,000 workers, commutes to the central 250 work days a year and two trips a day, and each hour is employment location. Each travels an average of 5 miles valued at $0.60, the added cost of travel time in this city per trip over congested streets at a speed of 20 miles per is $75 a year. To compensate, the wage in the large city hour (the speed would be 35 miles per hour if there were would have to be high enough to yield at least this much no traffic congestion). If half the new entrants commute additional income. to the center of the city (the same assumption as Next, consider the pollution effect. Particularly in the for the existing labor force), the average new entrant drier summertime, dust and other suspended particu- into the city's labor force makes five rush-hour trips per lates are an important source of pollution. Much of this week over this highway system. The effect on traffic is natural, but human activity also causes much of it, speeds and on commuting times for the other motorists and smaller cities are therefore generally less dusty than can be estimated through the use of the engineering larger ones. A typical level of suspended particulates relationship v = T - kq, where v is the speed under would be 150 micrograms a cubic meter of air, whereas congested conditions, u- is the speed under noncongested in smaller cities and towns a representative level would conditions (35 miles per hour), q is the traffic volume be perhaps 100 micrograms. Studies of the relation (75,000 cars per hour), and k is a constant (here, between property values and air quality, indicate that 2 .10-4), giving v equal to 20 miles per hour (the ex- individuals are willing to, and in fact do, pay higher rents isting speed noted above). Now recalculate v with to escape unfavorable air quality. There have been q = 75,001 cars per hour to estimate a new speed. numerous more direct estimates of the physical damage Subtract the old from the new speed to obtain the caused by pollution. A rough estimate of the annual cost change in speed. Divide this change into the 5-mile trip of pollution is $1 per microgram for a family of four on congested streets and multiply by the value of time, persons, a figure that reflects damage to health, short- $0.60 per hour, to obtain the added cost that an extra ened life spans, deterioration of durable property, and trip imposes on a trip taken by someone already in the 54 George S. Tolley city, $1.50 * 10-6 Multiply this result by 0.5, since the of employees are set through a political or negotiating commuting assumption implies that only one-half trip process, and situations in which foreign corporations per new worker is added to the congested streets. Then may decide for a number of reasons to pay more than the multiply by the 75,000 affected workers and by the 500 market-clearing wage. The wage in protected employ- work trips they each make per year (250 work days times ment is higher than elsewhere in the economy. two trips per day-home to work and work to home) to A simple, well-known condition that has been used to arrive at a yearly figure of $28 extra commuting cost that analyze protected employment is based on the hypoth- a new worker imposes on the population already in the esis that workers will distribute themselves so that the city. wage received in unprotected employment ends up The impact of the additional family on damages from being just as great as expected or prospective earnings in air pollution can be computed with the use of data protected employment, taking account of the greater already cited. An inference from observations on air probability of unemployment that workers face if pro- quality and metropolitan population size is that the tected employment is sought. People are willing to face average particulate level is likely to rise by 4 * 10-5 the unpleasant prospect of greater unemployment, micrograms a cubic meter as a result of adding one which requires drawing down savings or asking kin for family. If the damage value of $1 a microgram is applied largesse, in return for the hope of higher earnings if they and it is assumed that there are 150,000 families in the eventually succeed in finding a protected job. A queue of urban area, the overall damage value is 0.00004 times $1 people seeking protected jobs forms, and it grows until times 150,000, or $6. expected well-being in the protected employment sector Taking account of both congestion and pollution, the is driven down to that in unprotected employment. additional costs imposed on all other families when a In urban protected employment, which is not all family moves into an urban area of 600,000 population urban employment and may be only a small part of it, the amount to $28 plus $6, or $34 a year. probability of employment is sometimes approximated These results should be considered in relation to the as the ratio of total protected employment, nue, to the previous conclusion, that wages in this city would be total number of people seeking protected employment, about $125 higher than in other places of smaller size. n,. The rationale is that the total supply provides a pool When a family moves into the area, if there is a free from which those seeking employment have a chance of market equilibrium of city sizes generally, the personal being drawn. The wage condition is that the wage in gain is the $2,125 wage, for which $125 urban costs are unprotected employment equals the expected wage in incurred in this city and a $2,000 wage elsewhere is protected employment, which in tum is the proba- forgone. The other families in the large city suffer a bility of employment times the wage received if em- combined loss of about $34. Thus there is a net loss ployed in protected employment, giving wa = (nfue l of $34. nfu)wu. In terms of percentage changes the wage condi- Although the example is only suggestive, it gives a tion is i, = hue - nu + &i', which is the first of five perspective on the magnitude of the effects connected conditions for this model. with the physical environment. The $34 annual cost Meanwhile, the change in total employment in the neglected in migration decisions is 1.6 percent of the economy as a whole as a consequence of a rise in the wage rate. The demand for labor in any one city may be protected employment wage is the weighted sum of quite elastic owing to possibilities for undertaking pro- percentage changes in protected and unprotected em- duction elsewhere. For example, if the elasticity of de- ployment. This implies as a second condition n,h, = mand for labor is unity, the number of people in the city nhai2a + nue,ue* As in the earlier analysis, the weighted is made 1.6 percent larger by the neglected costs. If the sum of percentage changes in people seeking employ- elasticity is as high as 3, the city is made about 5 percent ment is zero, given the total population in the country. larger. This implies as a third condition nafla + nuhu = 0. Other conditions from the earlier analysis are that the Protected Employment percentage changes in employment equal the percent- age change in wage multiplied by the elasticity of de- A phenomenon that has received much attention is mand for labor. The fourth and fifth conditions are thus the attraction to cities of persons who are willing to h, = ,3awa and .ue = Pu1t'u- accept unemployment in return for being in a position The five conditions determine the percentage effects to obtain protected employment should the opportunity of &i, (the percentage excess of the protected employ- arise. Examples of protected employment include em- ment wage over the unprotected wage) on the unpro- ployment effectively covered by legally decreed mini- tected wage, total employment, unprotected employ- mum wages, govemment employment in which wages ment, protected employment, and numbers of people Market Failures as Bases of Urban Problems 55 seeking protected employment (u'0, ne,a, hue, and hu). from other urban employment, there is no net effect on The effect on the total number of people in urban areas is urbanization. given by the number of persons seeking protected em- The unemployment induced by wage protection, as a ployment. If that number increases, migration from percentage of total protected employment, is derived by rural to urban areas may be induced as people come to subtracting the percentage change in protected employ- cities in hopes of finding a job at higher real wages. ment from the percentage change in the number of Solution of the five conditions for the percentage change people seeking protected employment. It is n0 Inue times in numbers of people seeking protected employment the equation 4-2 coefficient plus the absolute value of gives the elasticity of demand for labor. Whether the unem- 1 + n ployment is substantial depends on the magnitude of the (4-2) nu = Mu * wage increase and on the amount of protected employ- (4-2) = 1 - flu lnaFa ment. The unemployment, again, may be limited. To the A conclusion apparent from this result is that the extent that workers rotate or are in the same families, existence of the higher wage may either raise or lower the unemployment does not affect the distribution of the total number of people who offer themselves for income. protected employment. The elasticity of demand for Still more refined models might allow for workers to labor in protected employment, 0, is negative, and if it be doing something else while they are looking for jobs. is greater than one in absolute value, the sign of the It seems possible that the refinements would lead to effect of the rise in protected wage as given by the even smaller estimated effects of wage protection on numerator of equation 4-2 is negative. With constant urbanization and unemployment. Reliable empirical factor shares and with other product and factor prices evidence is difficult to obtain because of the difficulty of unchanged, the absolute value of the elasticity of de- obtaining and interpreting unemployment statistics. mand for labor will be the reciprocal of labor's share and will thus be greater than one, which ensures a negative Benefits fom Government Expenditure effect. If product prices are raised, the absolute value of and Tax Policies the elasticity will be even greater. With an absolute value of elasticity in the vicinity of one, which is not too The benefits of such urban services as water supply, unlikely, the effect on the number of people seeking publicly provided health facilities, unemployment com- protected employment is about zero. The maximum pensation, social security, and maintenance of low food increase in the number of people seeking protected prices may make cities attractive if, on net, the charges employment is obtained in the unlikely event that the and taxes resulting from residence in urban areas are elasticity of demand for protected labor is zero and the less than the benefits. The net benefits of urban resi- elasticity of demand for unprotected labor is infinite. In dence that result from government actions can be ex- that case the coefficient in equation 4-2 reduces to one, pected to be reflected in a compensating wage differen- an indication that the increase in the number of people tial, along with the other differences between urban and seeking protected employment is proportional to the rural areas that go into determining urban-rural wage excess of the protected wage over the unprotected wage. differentials. Wages are likely to be raised substantially only in cases The role of the tax structure in determining net ben- in which the amount of protected employment is rather efits of urban residence deserves special analysis. Taxes small. Only in very unusual cases would there be a large imposed solely on the site value of land can be expected effect on wages in protected employment if the protected to be neutral with respect to location decisions, but taxes employment were a sizable fraction of urban employ- on land improvements or on capital structures, such as ment. Thus, even in the extreme case of zero elasticity of property taxes, may affect incentives to add to capital. If demand for labor, the effect of wage protection on the agricultural and urban property are taxed differently, the number of people in urban areas would be limited. It has location of capital may be affected. Since the owner of been shown that because of labor turnover, the elasticity nonhuman capital does not have to reside where the required to produce an increase in people seeking pro- capital is employed to collect the income, a condition for tected employment is even smaller than unity. The effect location neutrality is that the marginal tax rate on on the number of people in urban areas is further limited nonhuman capital be the same in different locations in by the fact that in the usual case, in which the protected money terms. The owner of human capital (the person wage applies to only a fraction of the workers in urban himself) must reside where the services are rendered. areas, the people seeking protected employment come Thus, a condition for location neutrality for human not only from rural areas but also from other employ- capital is that marginal tax rates on human capital in ment in urban areas, and to the extent that they come different places be the same in real terms. 56 George S. Tolley The urban wage tends to be lower than it would be such, and perhaps it is a greater concern. The reasons for, without net government urban benefits by the percent- tendencies toward concentration are in part the same age at which the net benefits are valued. This gives a first as those for urbanization in general; natural population condition that may be expressed in percentage terms as growth is a prime reason. Still, the question remains = = WiU + 9, where g is the percentage of the wage at why population should be so heavily concentrated in a which the government benefits are valued. That condi- few places. Perhaps the single most important reason for tion may be combined with three earlier conditions- tendencies toward large cities is that it is economical to that the total population is distributed either to rural or centralize production to avoid the costs of transporting to urban areas (n,aha + nun,, = 0) and that the changes goods back and forth. Only when these advantages are in employment are governed by the elasticities of de- overcome by rising costs of various kinds within a city mand for nonurban labor (hna = pa,a) and for urban does it become economical to decentralize. This is a labor (h, = Ptuj. Four conditions thus determine pervasive factor the world over, in developed and in percentage changes in urban and nonurban wages and developing countries. employment (tw, ,a, nt, and h) as a function of the A possible reason for the greater concentration in percentage of the wage at which government benefits developing than in developed countries is that the effect are valued, g. of the urbanization externalities considered above may Solution for the percentage increase in urban employ- be greater. In particular, net benefits from government ment yields actions could act to promote concentration in the _ P. largest cities. It would appear that the question of urban (4-3) hu = + g bias (if it exists) should be rephrased to deal not with 1 + n.u.uan3a exploitation of rural by urban people but rather with the Since the elasticity of demand for labor, P, is negative, effects of urban bias on city size. The existence of net the effect of government-supplied urban benefits is benefits in urban areas encourages people to move unambiguously positive. Furthermore, in contrast to toward those benefits. That movement dissipates tempo- the effects on urbanization (considered above) of en- rarily induced rural-urban differences in well-being but vironmental externalities and protected wages, it may have significant effects on city size. But although appears that government benefits could substantially uninternalized externalities may contribute somewhat affect the number of people in urban areas, at least on a to larger city size, it does not appear possible to make a one-time basis. For example, if the elasticity of demand really strong case that they are the prime cause of pro- in absolute terms is about 1 and if the proportions of nounced large urban concentrations in developing people in rural and in urban areas are about the same countries. It seems necessary to look elsewhere for a full (that is, the denominator is 2), the value of the coef- explanation. ficient in equation 4-3 will be about one-half. One reason Another hypothesis, which appears more likely, is that the effect may be substantial is that it applies to all intercity transport, which would permit decentraliza- residents of urban areas, on the assumption that none of tion to smaller cities, is a relatively expensive commod- them can be effectively excluded from government- ity in developing countries. Provision of basic road facili- supplied urban benefits. If these benefits are valued at 10 ties is highly capital-intensive, as is the operation of percent of the wage, the number of people in urban areas trucks and other transport equipment. If this hypothesis will increase by 5 percent. is valid, a gradual tendency toward less centralization may be observed as development becomes more com- plete and the price of human inputs rises in relation to the price of physical capital. This tendency toward a Primal Cities lowering of the relative price of intercity transport, however, may set in only at a late stage of development. In the developed countries there are signs of decen- A final, related, possibility is put forth more tenta- tralization not only to suburbs but to smaller towns and tively. Infrastructure decisions may not provide as ade- even rural areas. But developing countries are moving quately as would be beneficial for development outside toward even greater urban concentrations, evidence not large urban centers. Development of more adequate only of rapid overall urbanization but also of more pro- evaluation tools that would allow more accurate assess- nounced tendencies than in developed countries to have ments of infrastructure investments might lead to more a preponderance of population in one or a few cities. investment outside large cities. The tendency to concentrate population in one or a These hypotheses need to be tested as part of an few cities is as much a policy concern as urbanization as investigation of the fundamental determinants of city Market Failures as Bases of Urban Problems 57 size, as represented by the study by Henderson (chap- government transfer and price programs. With regard to ter 7). public services, there would be no externalities if recip- ients of the services were charged marginal costs and if strong attempts were made to carry provision of services Economic Policy and Urbanization to the point at which marginal cost equaled marginal benefits. Quite possibly, neither of these two conditions Most of the reasons for rapid urbanization in develop- can be fulfilled. In countries where public services are ing countries are closely related to the economic de- markedly more available in large urban areas than else- velopment process itself. Wholesale reversal of urbaniza- where and where the services are provided free or below tion is surely unthinkable and would amount effectively cost, raising the charges or extending the services more to a great interference with development, if not its vir- widely throughout the country are possible actions. Ex- tual denial. The more relevant question becomes tension of services may be the more realistic policy whether something marginal should be done to try to direction and may have distributional benefits. affect urbanization and in particular the concentration of urbanization in a few very large places. Infrastructure Decisions Externalities One method of providing infrastructure is to concen- trate facilities at places where demand for infrastructure Three types of policies toward urbanization may be services is growing, by mechanically extrapolating re- identified. The first is the internalization of externalities. cently observed trends in demand. The method is safe If pollution, congestion, and related environmental ex- because it has a self-fulfilling element: provision of in- ternalities increase with city size, as suggested earlier, frastructure services in an area encourages people to their internalization will lead to consideration of en- reside in that area. Demand then increases, and more vironmental costs in private decisions and will encour- infrastructure is built. Provision of infrastructure thus age decentralization. To affect decentralization, the feeds on itself. For example, road networks that make policies must distinguish between the severity of en- the larger cities ever more powerful hubs of transport vironmental costs in different places. For example, if may be progressively reinforced in response to the signal pollution abatement measures are tailored to local cir- of growing traffic brought on by previous investments cumstances, greater controls will be imposed where and further contribute to the large size of the cities. control costs are smaller and damages are greater. In At the opposite extreme, and less frequently, infra- environmental policy as now practiced, however, uni- structure may be provided to stimulate areas of lagging form environmental standards are often imposed across development or low population. The degree of success all or a large part of a nation, so that control require- may be disappointing because natural conditions and ments are the same irrespective of city size, and thus other locational characteristics are not overcome by the incentives to decentralize are thwarted. If environmen- new infrastructure, with the result that the infrastruc- tal controls are to play a role in decentralization policy, a ture is not used fully and the areas do not grow as much less procrustean approach (which would also contribute as was hoped. to the efficiency of environmental policy) is needed. Neither extreme appears optimal. Development of im- The earlier analysis suggests that the externalities proved methods of evaluation could have important con- connected with high wages in protected employment sequences for the spatial distribution of population. may either increase or decrease urbanization. The most More attention could be given to the effects of infra- important consequences may be unemployment and structure on the location of activity and to the quanti- overinvestment in search costs. It is not clear that wage tative estimation of benefits and costs that result from policy and urbanization policy as such are as closely alternative patterns of development. In particular, eval- connected as might appear at first sight. At any rate, uation techniques for transport facilities are sophis- public choice considerations make it difficult to elimi- ticated in some ways but need further development. For nate wage protection once it is instituted. A promising example, the travel savings method, which is widely approach may be to reduce the effects of protection by used to evaluate the benefits of road investments, is to be keeping annual nominal rises in protected wages low in recommended if the origin and destination of trips is not relation to the inflation rate. affected. More reliable methods, however, are needed to The analysis suggests that the largest opportunities estimate benefits if changes in trips are brought about for internalizing externalities may be in the provision of when road investments induce changes in the location public services, the tax structure, and the operation of of activities. Development of improved methods could 58 George S. Tolley lead to identification of alternatives that are preferable be commended, but this achievement so far remains to expansion in the largest cities; investments might be elusive. made, for example, in favorably situated medium-size Another kind of intervention is inevitable: the public cities. management decisions required to provide urban ser- vices, regulate and facilitate land use, and provide for financing. Various facets of these problems are dealt with in the studies by Bahl and Linn (chapter 10), Nearly every country has policies that try to alter Ingram (chapter 11), Pach6n (chapter 12), Thomas incentives to choose particular locations. Often such (chapter 13), and Keare (chapter 14). policies are explicitly concerned with preventing the Because public management decisions for cities go far concentration of the population in large cities. The poli- beyond spatial considerations, they are at once easier cies take many forms, including land use controls and and harder to deal with. A large body of ready-made subsidies. With notable exceptions, such as the studies knowledge and tools can be brought to bear on these by Lee (chapter 8) and Reif (chapter 9), few serious more traditional subjects, but there is a worldwide lag in attempts have been made to quantify the effects of the coordinating expertise and combining economics with policies. Many policies appear to be largely nominal, engineering, planning, law, and other disciplines re- without any real teeth, but in a few cases they may be quired to make public management decisions. Multiple overeffective. goals also make public management decisions difficult. Direct interventions may be decided on either because Income distribution considerations are difficult to deal the best economic policies, connected with externalities with at best and may run counter to spatial objectives, to or infrastructure, are not politically feasible or because say nothing of more general economic development deliberate actions to influence location, going beyond objectives. One of the hardest questions is how much externalities and infrastructure, are mandated-de- effort to put into city management decisions. The sirably or undesirably-by political considerations. In mainsprings of development are not found in many of either case, there are efficient ways to design the in- these decisions, but failure to deal with them adequately terventions. A Pigovian tax subsidy approach may be may put drags on development. Which ones will take mentioned as a way to achieve efficiency in interven- care of themselves somehow and which ones call for tions. It should be noted, however, that Pigovian special efforts are concerns for economic development. approaches-despite a century of advocacy-have sel- Yet there is a way to improve urban management at dom, if ever, been found to be amenable to practical present from a spatial point of view. It is quite practical application. Under a tax subsidy approach that is de- to introduce quantification of the value of externality signed to encourage a coherent amount of decentraliza- effects into the evaluation of projects. Orders of magni- tion, a shadow price for additions to population in differ- tude of the value of externalities at the margin can be ent places, such as in cities of different sizes, would be developed along the lines illustrated in this chapter. established. Firms would face subsidies or taxes accord- These can be used as add-ons in social benefit-cost deci- ing to the amount of their employment in different sions concerning roads, water projects, and other ex- places, and the tax or subsidy would be proportional to penditures that affect the location of activity. In this the difference in shadow price of having people in differ- way, spatial externalities could be taken into account in ent places. Alternative forms of inducements (such as the workings of government, without adopting explicit interest-rate and energy subsidies) and quantitative con- spatial policies. The required quantification would prob- trols (such as land use restrictions) would be judged ably reveal the externalities to be small in most cases. according to how well they approximated the outcome Quantification would help keep externalities from being that would be attained under the shadow price approach. blown out of proportion by the emotional beliefs and The same standard could be used to design and evaluate political distribution motives that too often figure in more general overall strategies for spatial development. discussions of externalities. Deliberate spatial interventions as described above are controversial. Realistically, in the absence of badly needed objective quantifications of the effects of spatial intervention policies, it is difficult to be optimistic that Bibliography effective or desirable interventions will occur soon. Efforts to design interventions rationally and even bring Brueckner, Jan K. 1981. "Labor Mobility and the Incidence of in cultural diversity and national defense as spatial con- the Residential Property Tax."Journalof Urban Economics, siderations, along with more usual externalities, are to vol. 10, no. 2 (September), pp. 173-82. Market Failures as Bases of Urban Problems 59 Chappie, Mike, and Lester Lave. 1982. "The Health Effects of . 1982. "Evaluating Consumer Amenities and Interre- Air Pollution: A Reanalysis." Journal of Urban Economics, gional Welfare Differences." Journal of Urban Economics, vol. 12, no. 3 (November), pp. 346-76. vol. 11, no. 1 (January), pp. 32-59. Cooley, Thomas F., and C. J. La Civita. 1982. "A Theory of _ . 1982. "The Impact of Govemment Policies on Urban Growth Controls."Journalof UrbanEconomics, vol. 12, no. Concentration." Journal of Urban Economics, vol. 12, 2 (September), pp. 129-45. no. 3 (November), pp. 280-303. Cropper, M. L. 1981. "The Value of Urban Amenities."Journal - . 1985. "Property Tax Incidence with a Public Sector." of Regional Science, vol. 21, pp. 359-74. JournalofPoliticalEconomy, vol.93 (August), pp. 648-65. Cropper, M. L., and A. S. Arriaga-Salinas. 1980. "Intercity . 1986. "Efficiency of Resource Usage and City Size." Wage Differentials and the Value of Air Quality." Journal of Journal of Urban Economics, vol. 19, no. 1 (January), pp. Urban Economics, vol. 8, no. 2 (September), pp. 236-54. 47-70. Diamond, Douglas B., Jr., and George S. Tolley, eds. 1982. The Izraeli, Oded. 1977. "Differentials in Nominal Wages and Economics of Urban Amenities. New York: Academic Press. Prices between Cities." Urban Studies, vol. 14 (October), Fare, Rolf, Shawna Groskopf, and Bong Joon Yoon. 1982. "A pp. 275-90. Theoretical and Empirical Analysis of the Highway Speed- . 1983. "Federal Individual Income Tax and Allocation Volume Relationship."JoumalofUrbanEconomics,vol. 12, of Labor between sMsAs." Journal of Regional Science, vol. no. 1 (July), pp. 115-21. 23, pp. 105-13. Fogarty, Michael S., and Gaspar Garofalo. 1980. "Urban Size McConnell, Virginia D., and Mahlon Straszheim. 1982. "Auto and the Amenity Structure of Cities." Journal of Urban Pollution and Congestion in an Urban Model: An Analysis of Economics, vol. 8, pp. 350-61. Alternative Strategies." Journal of Urban Economics, vol. Graves, Philip E. 1980. "Migration and Climate." Journal of 11, no. 1 (January), pp. 11-31. Regional Science, vol. 20, no. 3 (November), pp. 227-37. McDonald, John F. 1983. "An Economic Analysis of Local - . 1983. "Migration with a Composite Amenity: The Role Inducements for Business." Journal of Urban Economics, of Rents." Journal of Regional Science, vol. 23, vol. 13, no. 3 (May), pp. 322-36. pp. 54146. Porell, Frank W. 1982. "Intermetropolitan Migration and Grieson, Ronald E., and Richard Arnott. 1981. "Optimal Fiscal Quality of Life." Journal of Regional Science, vol. 22, pp. Policy for a State or Local Government." Journal of Urban 137-58. Economics, vol. 9, no. 1 (January), pp. 23-48. Power, Thomas M. 1981. "Urban Size (Dis)amenities Revis- Harris, John R., and Michael P. Todaro. 1970. "Migration, ited." Journal of Urban Economics, vol. 9, no. 1 (January), Unemployment, and Development: A Two-Sector Analysis." pp. 85-89. American Economic Review, vol. 60 (March), pp. 126-42. Rosen, Sherwin. 1979. "Wage-Based Indexes of Urban Quality Haurin, Donald R. 1981. "Local Income Taxation in an Urban of Life." In P. Mieszkowski and M. Straszheim, eds., Current Area." Journal of Urban Economics, vol. 10, no. 3 (No- Issues in Urban Economics. Baltimore, Md.: Johns Hopkins vember), pp. 323-37. University Press. - 1983. "The Effect of Property Taxes on Urban Areas." Tolley, George S. 1974. "The Welfare Economics of City Big- Journal of Urban Economics, vol. 7, no. 3. (May), pp. 384- ness." Journal of Urban Economics, vol. 1, no. 3 (May), pp. 96. 32445. - 1984. "The Effects of Output and Factor Subsidies or Tolley, George S., Philip E. Graves, and John L. Gardner, eds. Taxes on an Urban Area." Regional Science and Urban 1979. Urban Growth Policy in a Market Economy. New Economics, vol. 14, pp. 533-46. York: Academic Press. Henderson, J. Vemon. 1977. Economic Theory and the Cities. New York: Academic Press. 5 Urban Development Policies in Developing Countries Bertrand Renaud The recent world recession-some might even call it has been sweeping through developing countries since T a depression-is testing the vitality of urban poli- World War 1I will be sustained through the end of this cies in developing countries. One is reminded of Dick- century. Second, there are significant quantitative dif- ens's Tale of Two Cities: "It was the best of times, it was ferences between contemporary urbanization in de- the worst of times." On the one hand, the pressures to veloping countries and the historical patterns observed move the economy again, to confront the problems of for today's advanced economies. Third, the policy prob- sluggish exports, and to lower high unemployment lems encountered in developed countries have clearly rates, together with the lack of government resources been diverging from those of developing countries. Sim- for state and local investments, make it the worst of ple extrapolations of past trends in developed countries times for urban development policies. On the other will be of limited value for the formulation of urban hand, it is generally in periods of strain or crisis that the policies for developing countries. significance of the urban sector for the national econ- omy is most truly appreciated. Austerity is forcing peo- The Main Characteristics of Urbanization ple to rethink the fundamental issues of urban develop- DevMam Conries ment and the types of policies that can be most fruitfully in Developing Countres pursued. This may be the best of times to review policy There were approximately 800 million people in 1750, priorities and to look for robust solutions to urban about 1.3 billion by 1850, 2.5 billion by 1950, and over problems. 3.6 billion in 1970; there will be between 6.2 and 6.5 Before the prospects for urban development policies billion people by 2000. Most of this population growth is are outlined, some questions must be answered. What taking place in developing countries. For instance, the are the characteristics of policies toward urbanization 1982 census shows that the population of China is more and the policy tools that have been used in developing than 1 billion and is growing by 17 million a year. These countries? What rationales are given for those policies? demographic changes have been associated with a rapid What do we know about the effects and effectiveness of redistribution of population from rural to urban areas policies? Without answers to such questions it is not and a significant shift in the proportion of the urban possible to examine properly the future direction of population that lives in large or very large cities. Before urban development policies. 1850 no country was predominantly urban, and in 1900 only the British population was more than 50 percent urban. By 1920, 14 percent of world population was Basic Trends in Urban Policy urban. Now that proportion is close to 40 percent, and it is anticipated that by 2000 half of the world's population The forces that shape urbanization and urban policies will be urban. The urban areas of developing countries in developing countries can be summarized in three absorbed about 600 million people between 1950 and broad propositions. First, the wave of urbanization that 1980. During the final quarter of this century there are 60 Urban Development Policies in Developing Countries 61 expected to be 1 billion people in developing-country million. In 1975 the corresponding figures were 983 cities, which is 64 percent more than the total popula- million and 51.4 percent. For 2000 some projections tion of about 610 million in the advanced market econo- yield 2.2 billion and 63.0 percent. mies today. * The striking growth of urban areas should not dis- Demographic and economic forces render contem- tract attention from two facts: many countries have very porary urbanization different from the earlier experi- large rural populations, and a very large proportion of ence of advanced economies. First, population in de- the urban population still lives in towns of less than veloping countries is expanding rapidly. A country that 100,000 population. Because very large cities dominate grows at the rate of 2.4 percent a year, the current group the urban systems, there is a great difference between average for middle-income countries, will experience an the statistical average size of settlements and the typical 81 percent increase in population in twenty-five years. size of settlements. There are very large urban places, Second, the middle-income countries have experienced but the typical place is still rural oriented. high economic growth rates so far. After twenty-five * Even though developing countries are urbanizing years an economy that is growing at 5.5 percent a year- rapidly, only a few have declining rural populations. This the GDP-weighted average growth rate for the middle- is because of their high rate of population expansion- income countries for 1970-79-will be 3.8 times larger the average crude growth rate is 2.4 percent a year, than it is today. Rapid demographic and economic about five times that of industrial countries, which reg- growth contributes to average growth of 3.8 percent a ister an average 0.4 percent a year. The reservoir of year in the urban population; in many countries the potential rural-to-urban migrants continues to expand. growth rate is much higher. Institutionally, the role of government is more important in developing countries The evidence suggests that the policy context in de- now than it was for advanced economies when they were veloping countries differs markedly from the better- at similar levels of urbanization. The slowdown of the known historical pattern of Western countries. In in- world economy in the 1980s will reduce the pace of dustrial countries urbanization took many decades and urbanization in developing countries. Nevertheless, the occurred relatively slowly in comparison with the urban historical contrast remains sharp between the slow transformation now occurring in developing countries. urbanization at relatively high levels of income that In today's developing countries the pace is more rapid, advanced economies experienced in the past and the population growth is higher, income levels are lower, rapid urban population concentration at much lower and the opportunities to relieve domestic population levels of income in developing countries today. pressures through migration are limited. World urban trends can be described further in a series of propositions. - Diverging Urban Trends * Urban areas will play an increasing role in absorbing in Developing and Developed Countries large shares of the world's population. There will be a Since the early 1970s urban trends have been moving marked increase in the level of urbanization over the in divergent directions for developed and developing next twenty years. An increasingly large number of countries. It is worthwhile to briefly compare these countries will have become more urban than rural. tendencies because of the influence that urban policies (Already, more countries are predominantly urban than and analyses in developed economies have had so far on are predominantly rural.) the formulation of urban policies in developing coun- * The distribution of urban population over broad tries. The contrast between the two groups of countries regions of the world has changed dramatically. At the may facilitate the discussion of urban policies in de- beginning of this century the largest share of the world's veloping countries. population living in cities of over 100,000 was found in The difference between urban conditions in develop- Europe. By 1950 Europe was already behind Asia and ing countries and in advanced economies is most easily America, and by 2000 Asia will dominate the world brought out by reducing the complex socioeconomic urban picture, with about 45 percent of the world's processes of urbanization to a single demographic in- urban population. The level of urbanization will rise dicator, the percentage of total population that is urban. most rapidly in Africa. Long-term urbanization, which involves an increase in * There is a trend toward concentration of urban this percentage, can then be seen as the process of population in large cities (over 1 million), very large moving along a logistic curve, with the urban share of cities (over 2 million), and supercities (over 5 million). total population rising from about 1 or 2 percent to a In 1950 the world's urban population was 393 million, saturation level somewhat less than 100 percent. The and 45.1 percent of that number was in cities of over 1 dynamics of urbanization in a given country can be 62 Bertrand Renaud characterized by the slope of the urban logistic curve, the third industrial revolution in the advanced econo- which reflects the tempo of urbanization, and by the mies. The first industrial revolution, which started ev- urban saturation level, defined as the highest percentage erything, was based on coal, the steam engine, railroads, of urban population that is likely to be urbanized at very and textiles. The second industrial revolution was based high levels of economic development.' In comparing on petroleum products, chemicals, the automobile, and national urbanization strategies, much confusion could electrical and mechanical industries. The current rev- be avoided by referring to this simple and familiar curve olution is based on new energies, electronics, informa- as a quick way of differentiating between countries, tion industries, bioengineering, and services. The first since the level of urbanization and the level of economic two revolutions, especially the first one, strongly favored and institutional development are closely correlated.2 urban concentration; the third one does not. According Until the 1960s advanced and developing countries to this interpretation one could hypothesize that the alike were seen as moving along this urbanization curve, 1980s are not the beginning of a period of permanent with the first group preceding the second. In the 1970s slow growth but rather a transition period toward new two major structural changes became apparent in de- forms of growth. The policy implications of these two veloped countries, and their impact on urban policies is structural changes for urban policies in developed coun- becoming pronounced. The first structural change is the tries are becoming clear. end of traditional urbanization: the national populations Until the late 1960s the policy paradigm in advanced of advanced economies have become fully urbanized, economies was based on urban growth under these con- rural-to-urban migration has slowed to a trickle, fertility ditions: a moderately rising national population; con- rates in many countries are falling below the replace- tinuing transfers of population from the countryside ment level, and the populations of some countries are (associated with the development of manufacturing and even declining. The steadying effect of internal migra- service employment); a differential growth that favored tion on the growth of most cities is now missing in many larger cities and the formation of city regions; an in- advanced countries. In addition, national governments creased demand for space per capita (associated with are trying to limit international migration flows. Urban significant gains in per capita income); and a long-term systems such as those of the United Kingdom and the concentration of population in favored economic re- Federal Republic of Germany have stopped expanding. gions. Demographic urban growth often takes on the appear- Confronted with urban growth, advanced countries ance of a zero-sum game in which the demographic developed policies incrementally and responded to prob- growth of one city comes more and more at the expense lems as they emerged. At first, only limited attention was of another city, whether it is defined as a central city, paid to the context within which urban issues were suburban center, metropolitan area, or medium-size or debated and evaluated. Then rapid urban growth, com- small town. For the first time advanced countries are bined with abundant fiscal resources, led to an increase discovering what it means to have reached the urban in the scope of objectives, in the breadth of the popula- saturation level at the end of the logistic curve. tion to be served, and in the number and variety of The other major structural change is economic rather instruments to be tried. A major shift occurred, from an than demographic. It is marked by the transition from early and almost exclusive consideration of physical an extended period of rapid and sustained growth during planning problems to a more comprehensive focus on the 1950s and 1960s to the present period of slow and the socioeconomic problems of cities. Increasing em- uncertain growth and is tied to the emergence of a world phasis on income redistribution in an urban context was market over the past thirty years, the accompanying accompanied by a concern for economic efficiency in shifts in the international division of labor, and, in solving problems. The rapid growth of urban programs particular, the impact of these shifts on the demand for frequently led to problems of coordination and and production of standardized manufactured goods. bureaucratic congestion. During this period parallels During the 1970s the triple impact of the slowdown in were drawn between urban policy issues in developed economic growth, structural change in the demand for and developing countries. manufactured goods, and shifting terms of trade with The factors that now differentiate urban policy issues the oil-producing countries accelerated the pace of in advanced economies from those of developing coun- change of urban policies in advanced economies, partly tries are many: the stagnation or decline of the largest but not exclusively because of the concentration of old- metropolitan regions in developed countries, greater line manufacturing activities in certain regions and inequalities within large urban areas than between re- cities. gions in developed countries, a shift of manufacturing A stylized way to describe this economic structural activities in advanced economies from the production of change is to say that we are witnessing the beginning of standardized goods toward the production of goods with Urban Development Policies in Developing Countries 63 a high proportion of services input, and the dominant within large cities, have prompted policymakers to ex- role of multilocational, multifunctional corporations periment with spatial considerations in their national that do not simply react to their spatial and economic economic policy schemes. For example, they have be- environment but attempt to shape it. come more concemed with the composition of invest- In this new context of stabilizing urban systems and ment projects and their differentiated impacts on the fiscal retrenchment, urban policies and instruments in growth of regions and cities. developed countries will differ considerably from those of developing countries. In a surprisingly large number Ob-etives of Westem countries the policy paradigm is shifting Oec from urban growth to an emerging urban zero-sum In many countries, particularly middle-income coun- game. Because of the nationwide impact of industrial tries, the decentralization of economic activity from restructuring, which affects social groups selectively, large, congested capital regions and the closing of the the relative balance between social (people-oriented) welfare gap between rich and poor regions are the stated policies and urban (place-oriented) policies is shifting in policy objectives. The justifications for initiating urban favor of the former in developed countries. New urban decentralization policies generally given by developing- policies for developed countries must now be defined country govemments are, first, the necessity of reducing against four major issues: the severe managerial problems and the economic and . How tipvtecocasocial costs associated with congestion, pollution, and 1. Hwhic toniroventhe ejonprobmibse of reciiers,any the difficulty of providing adequate services in large of which confront major problems of reconversion .. cities; second, the need to reduce wide regional income from an old manufacturing base to services and disparities within the country; and third, the belief that high-technology activities more investment outside the largest urban centers in 2. How to improve social conditions, since industrial favor of the poorer outlying regions will foster national restructuring is leading to sharp income inequali- growth. Thus, urban decentralization policies are de- ties within cities. fended as both equitable and economically efficient. The 3. How to improve urban services to accommodate national urbanization policies of many developing coun- the changing size and composition of resident tries are based on the premise that it is possible and right populations while meeting major maintenance to decentralize population away from the largest cities. needs A national population decentralization strategy 4. How to allocate resources and responsibilities actually covers many objectives. This multiplicity of among levels of governments, given the heter- objectives calls into use a comparable multiplicity of ogeneity of local conditions among cities. instruments, and the problem becomes one of identify- To manage simultaneous urban growth and de- ing the combination of policies most appropriate for a line, the emphasis should be on selectivity, flexibility, given country at a given level of development. smaller-scale interventions, greater diversity of actions, Some of the more explicit objectives encountered in and increasing reliance on local government to coordi- developing countries are: nate activities at the city or neighborhood level. For the 1. The integration of peripheral regions to increase advanced economies there also remains the problem of the size of national domestic markets and the intensity declining cities, underutilized infrastructure, and the of regional demand. Raising regional incomes is greatly role of the central govemment in influencing such dependent on rural development, agricultural policies, trends. and the development of transport and communication networks. 2. The integration of peripheral regions and the Urbanization Policies opening up of new resources to raise national output. This objective applies to mineral-rich countries as well The rapidly shifting patterns of population distribu- as to countries which have had an important resource tion and economic activity in most developing countries frontier to exploit, such as Malaysia and Brazil. create inequalities in economic growth rates, industrial 3. The reduction of interregional disparities. In a structure, employment conditions, household incomes, well-integrated society, concern about inequality among wages, and levels of services which are keenly felt by individuals and households should be more important decisionmakers and social groups. These inequalities, than concern about inequality between places. The re- which include those between the rural and urban sec- duction of regional disparities is, however, a legitimate tors, regions, cities of different sizes, and social groups concern in developing countries. In Latin America, for 64 Bertrand Renaud instance, the output per capita of one region may be as structured way that has typified the experience of ad- much as ten times the output per capita of another, vanced countries during their decades of rapid urban while in fully integrated advanced economies the appar- growth. In fact, they have been derived from experi- ent differences are on the order of three or two to one. ments in advanced countries. 4. The improvement of national political integration These policies stress the decentralization of economic and social cohesion within the nation. It would be a activity and population as a means of relieving conges- conspicuous mistake to assume that national policies tion and solving the environmental problems of large have purely economic objectives such as increasing out- cities. Common objectives are to create employment and put or redistributing economic opportunities. In heter- service centers, to build new towns, and to implement ogeneous societies marked by important cultural, polit- large investment programs in transport and infrastruc- ical, and linguistic differences, a dominant element of a ture. Technical interest in the design of new facilities national urbanization strategy is to maintain the cohe- and new urban technologies, however, often overtakes sion of the state and to prevent regional minorities from, the understanding of the effective demand for services in some fashion, leaving the national coalition of social by urban groups. This supply-side, design-dominated groups. Special institutions such as regional develop- approach to urban development is only slowly being ment corporations are often used as an expression of the corrected through a better understanding of the en- central government's concern for the region. gineering, economic, and institutional constraints on 5. The rapid development of border regions for urban development at low levels of income. reasons of national security. This objective has been At the metropolitan level the overwhelming preoc- important in Latin America, where disputes over the cupation is with the absorption of a large population exact location of national boundaries in underdeveloped expansion every year. The supercities with populations regions have been frequent in the past. Similar situa- above 5 million may no longer be the fastest growing, tions also exist in Asia and the Middle East. and the current economic slowdown may marginally 6. The impro?vement of thenational systemof citie-s. reduce their growth rates. Still, they are likely to grow at This is a more recently stated objective of national poli- rates between 3 and 6 percent, or by 150,000 to 300,000 hies is more-recoen stae obec. Tie of national p a year. Most metropolitan governments have great dif- cies in middle-income countries. The city system has a dominant role in the transmission of economic impulses ficulty in developing and implementing well-structured and the diffusion of social and economic innovations investment programs to confront such problems. Urban over the national territory. Raising the level of economic areas are outgrowing their governments' capabilities. ove th naioa tertr. Rasn the... level of L......ocal governments have inadequate authority to coor- activity in medium-size cities will help them attain dnat activities beyond their own boundaries and their economies of scale and will improve the interregional d diffusion of economic growth. It is also expected that a political power and resources are inadequate for project more developed system of cities will help in equalizing implementation. Interventions by local governments accesstoimportantservices such as education and pub- need the support of central governments. New local tax lic health which can increase the human capital of a bases must be created and existing ones redesigned. cohathca Increasing reliance on transfers from central govern- country. ments seems desirable but is not necessarily feasible. The issues are familiar and can be listed briefly: Some Questions The above objectives are still too broadly defined. In * How to improve the economic base of the city and particular, the objectives of decentralizing population provide productive employment to a rapidly expanding away from the largest urban centers and narrowing labor force that often has low skill and educational interregional differences beg many questions. Can an levels? How to influence employment location to ease urban strategy operate independently of other national, the growth of the city? social, and economic strategies? Is decentralization * How to provide transport that is affordable for the really desirable from the viewpoint of economic ef- majority of the population? How to ensure a dynamic ficiency? Is there a tradeoff or a convergence between balance between the public and the private sectors when higher rates of national growth and greater regional a 10 percent minority of car owners can and does impose equality? Is it possible to channel economic activities to severe congestion costs on the other 90 percent of daily preselected cities? Under what conditions? At what cost? commuters by overwhelming an inadequate infrastruc- ture? Urban Problemns and Policies * How to develop land use patterns that generate a Urban policies in developing countries have been better balance between home and workplace? How to formulated and tried in the same incremental and un- recapture some of the increment in land values gener- Urban Development Policies in Developing Countries 65 ated by urban growth to finance city infrastructure in- good understanding of the unintended spatial effects of vestments? national economic policies-of implicit spatial poli- * How to improve housing supply when 90 percent of cies-is essential for developing countries. First, these the housing stock is provided by the private sector with- policies have a definite impact on where people live and out the necessary urban infrastructure which would work. Second, there is a growing awareness and specific make new neighborhoods fully functional? Put another evidence that the effects of implicit incentives on busi- way, how to take advantage of low-income progressive ness location decisions are much stronger than the housing investment patterns to develop affordable but publicized explicit incentives that favor decentralized efficient and healthy neighborhoods? location. Third, if national economic policies were ad- * How to mobilize financial resources to meet the justed so that their spatial bias in favor of the most massive infrastructure needs of cities for road networks, advanced regions were moderated, reliance on the in- transport systems, water and sanitation systems, educa- visible hand of the market would reduce the need for tion, local health services, safety, and fire protection? specialized teams of spatial planners and avoid the waste of scarce managerial and administrative talent on the whoare Howsto prov ede assistance tolteast rblesigents i fine-tuning of regional decentralization incentives of doubtful effectiveness or efficiency. The likelihood of As noted earlier, rapid demographic growth at low such an adjustment is very remote, however, because levels of income is characteristic of urban development coordination of the execution of better-designed policies in developing countries. Because of severe resource con- presents great problems. Here we are mostly interested straints, resource mobilization is as important as the in understanding the urban impacts of national policies distributive aspects of urban growth: resources must be and the dynamics of urban development in the manner found before they can be allocated equitably. In this new defined by Kelley and Williamson in chapter 3. period of severe capital scarcity, it is clear that to receive Regional policies can strengthen promising sec- support from central government planners, urban in- ondary urban centers through such actions as better vestments will have to contribute demonstrably to in- investment and management policies for transport, in- creasing the productivity of cities. This is a new chal- dustrial estates policies, and more important, the sys- lenge for urban analysts. How can it be shown that a tematic development of organized informational net- certain urban investment package raises the productiv- works between these cities and the capital region, such ity of a city? How can such a package be defined? What as banking networks, industrial association networks, are the analytical methods available for establishing this and better administrative structures. As is now well link? Can they be implemented easily and routinely? established, growth center strategies are more a way of thinking about such policies than a precise meth- Three Dimensions of Urbanization Policies odology. The appropriate internal management of cities is Urbanization policies can be viewed along three important to the success of national spatial policies. In dimensions. First are national economic policies which the case of very large cities, policies to limit or stop have unintended effects on the urban development of a population growth are not substitutes for policies that country. Second are regional policies that operate at the directly address the correction of congestion and pollu- national level and aim at an equitable and efficient tion and the provision of adequate services. If other allocation of population and resources among regions. cities are not efficiently and effectively managed, their Third are policies for the internal management of cities. chances of attracting industries and migrants from the The unintended spatial biases of national economic largest urban centers will be small. policies in favor of some urban centers are commonly products of trade policies which protect the manufactur- ing sector. For example, policies regarding credit alloca- Analytical Issues tion, public investment, and prices give preferential treatment to economic activities which are concentrated Progress in the formulation of urbanization policy is in a few cities and regions. The management practices of possible by integrating more closely the three dimen- the central government and its regulation of economic sions of policy discussed above. Research findings, activities require location of these activities close to the however, lag far behind the needs of policymakers. De- capital and contribute to the urban vortex. signing workable programs to moderate urban concen- It is not necessary to make the unrealistic claim that tration, to narrow the gap between rich and poor spatial considerations should prevail over economic regions, and to address more effectively the needs of growth considerations in the formulation of national lower-income groups strains our knowledge of national economic policies. But there are several reasons why a urban development processes. 66 Bertrand Renaud Much urbanization policy in developing countries and Figure 5-2. The Equity Rationale of National elsewhere is based on presumptions rather than on Spatial Policies established findings. For instance, there is always much interest in coordinated regional schemes in developing Price countries, but the actual motivation is the perception X M that trickling-down does not work fast enough to reach Supply curve poor regions and, particularly, poor people. It is argued S curve that, on equity grounds alone, it is necessary to increase the level of investment in peripheral regions, or at least T in nonmetropolitan areas within these regions. In addi-v tion, it is felt that both equity and economic efficiency N S ' Social demand curve objectives can be met jointly by investing in peripheral regions because such investment opens up new re- Private demand curve sources to the national economy. O Quantity The justification of national spatial policies on equity 0 Q R grounds is not a problem. It is accepted that public policy should help achieve a certain minimum standard of living in every region of a country. The minimum is offset by the loss in economic efficiency. Inherent in standard of living could be defined on the basis of the such spatial redistributive objectives is a certain amount market demand curve of a given percentile of the na- of efficiency loss, which is accepted by society on equity tional population, say the twentieth percentile, for im- grounds. portant goods and services such as housing, medical A major research task related to urbanization policy care, nutrition, and public education. For instance, in is the development of a methodology to determine figure 5-1 the private demand curve for nutrition leads whether the increasing concentration of population and to an equilibrium level of consumption equal to OQ in a economic development in a limited number of very large peripheral region. National intervention is considered cities in developing countries is a result of faulty policies justified to bring that level up to at least OR for all social and market failure, or whether rapid urban concentra- groups. The net social benefit of a transfer policy will be tion is an inevitable part of the development process. We a triangle MNT which is the difference between MNST, need to know whether decentralization policies can be the public good externality generated by providing justified on economic efficiency grounds, or whether everyone a minimum standard of consumption up to they are futile and economically inefficient. OR, and NST, which is the usual measure of the effi- ciency cost of providing a subsidy of up to ST to the target group. Exactly the same analysis is repeated in a The Direction of Urbanization Policies more general form in figure 5-2, in which the shaded areaMNSU represents the social gain from the external- Rapid urban growth will continue to shape policies in ity created by the regional transfer policy and the trian- developing countries in the coming years. The two tra- gle NST represents the gain from the externality which ditionally dominant urban issues-decentralization away from large urban centers and reduction of in- equalities between regions-will continue to shape the Figure 5-1. The Private Demand Curve forNutrition agenda. It is reasonable, however, to expect a more knowledgeable selection of urban development objec- Price tives and a greater degree of attention to the choice of instruments to achieve more realistic goals than in the past. Over the past decade the gap between the rhetoric '%,M Social demand curve of urbanization policies intended to prevent urban migration and the actual urban growth patterns (which \\ T Supply curve include rapid concentration in major cities) has become N \ Supply curve painfully clear. Everyone hopes not to hear again the old s question of whether urban growth can be stopped and I \ would rather be asked what kind of urban growth should | \Private demand curve be encouraged. Misleading notions of optimum geogra- I Level of nutrition phy and optimal size distribution of cities based solely on 0 Q R population should not be used to select population Urban Development Policies in Developing Countries 67 growth targets for which no feasible program can be they could hope to handle. The consequence has been found. hypertrophy at the center and atrophy at the local level. Significant progress has been made in understanding There has been significant deterioration of services, the the impact of the spatial distribution of urban growth as two most common examples being national housing a result of growth policies, trade regimes, economic corporations and water systems which are managed regulations, the influence of federal systems of govern- directly by ministries. Overcentralization has also af- ment, and fiscal policies. International comparisons also fected the financial health of cities, where a common show how ill-adapted earlier policy advice has often been error in English-speaking countries has been the as- to the structure of existing institutions within a country. sumption by the central government of revenue bases The World Bank, through its activities in about sixty which had historically been available to cities. In addi- countries, has gained substantial insights into the dy- tion, the financial responsibilities of cities, particularly namics of urban growth in developing countries and in education and health, have increased drastically, but into the actual capacity of national and local institutions new revenue bases have not been established. to induce change. In addition, there is better under- There has also been a systematic failure to evaluate standing of the time dimensions of urbanization policy. urban income distributions and to plan for them. Before A sobering experience for policy analysts with limited or independence most cities had high employment rates no operational experience has been the sudden aware- and few poor families, since urban migration was con- ness that most urban projects take an average of eight ditional on the availability of a job. In the postindepen- years from original planning to full-scale operation. dence period urban policies failed to plan for services Thus, urban project cycles are considerably longer than that the incoming migrants could afford. Thus, achiev- the political cycles in many countries. The momentum ing realistic living standards has been a pervasive prob- behind new urbanization policies depends on market lem. New urban migrants do not have a realistic sense of forces and the parallel growth of specialized urban in- what is economically possible, and there has been a stitutions that are less sensitive to short-term politics. strong political demand for high standards of services There are major differences in urbanization policies that are unattainable at present income levels. The around the world, and it is to these policies that we now typical result has been high-quality services for a few and turn. no services for most. This in turn has led to serious physical problems of unserviced and sometimes unser- Urbanization Policies in Sub-Saharan Ahica viceable squatter settlements, as well as overcrowding in Urbanization Policies in Sub-Saharan Africa pandadupandaes planned and unplanned areas. The severity and complexity of the problems facing This extreme polarity in income distribution has been many countries of Sub-Saharan Africa in the 1980s will maintained through the growth of public sector employ- greatly constrain urbanization policies in the region. ment, which inspired the formulation of the Harris- Because it is starting from the lowest urban base, Sub- Todaro model of rural-to-urban migration. Whereas for- Saharan Africa is the region that is undergoing the mal-sector employment accounts for 6 to 15 percent of greatest change, and its postindependence period has the working population, public-sector employment in- seen spectacular urban growth. The average growth of creased during the postindependence period and now urban population has been 6 percent a year-8.5 percent constitutes 40 to 75 percent of salaried employment.4 for thirty-five major capitals that are doubling their Not surprisingly, political problems have arisen in the populations every nine years. There are now twenty- day-to-day financing of cities. In most countries it has eight African cities with populations of over 500,000; been difficult to levy adequate charges for the sustained twenty years ago there were only three (World Bank development of such urban services as water supply, 1981). Because of this rapid growth, cities which were in waste disposal, electricity, and road maintenance. In- good physical and financial shape two decades ago now adequate charges have led to lagging and deteriorating face enormous difficulties. services and have thereby increased the resistance to The policy problems that are besetting African coun- paying for them. Current flat or negative growth rates of tries in the postindependence period are severe, and the gross national product, coupled with a projected annual same problems can be identified almost everywhere.3 population growth of 3 percent, are having a sobering There has been a conspicuous failure to predict and plan impact on urbanization policies in Africa. for rapid urban growth. Most city governments have Nigeria, the largest country in the region, well illus- been unable to grasp the implications of a population trates present problems. Its position in 1973 as a poor that doubles every nine years. Overcentralization has overpopulated country with a per capita income of $150 compounded the problems, as many new central institu- changed dramatically as oil revenues brought in an tions have assumed or claimed more responsibility than estimated $100 billion in foreign exchange earnings 68 Bertrand Renaud during 1973-81. But despite substantial petroleum rev- foundations for effective management in large and enues, exports, and a reported per capita GDP of $670 in medium-size cities. Given the scarcity of central re- 1981-as against the regional average of $410-Nigeria sources, a high-priority item is the development of a still ranks among the world's least developed countries. local tax base through the implementation of a cadastre Its urbanization has been greatly accelerated for the and an effective system of land administration. Policies worse by the oil revenues. Nigeria also provides a clear less biased in favor of the largest urban centers-in example of the dominant impact of trade and industrial particular, agricultural price policies-are more likely policies on urban growth and concentration. An analysis to be implemented now than in the past two decades. It by Bertrand and Robertson (1978) used the relatively is less clear whether trade and industrial policies will be standard estimation of the sectoral net subsidies pro- modified in a way that will favor secondary urban cen- vided by trade and industrial incentives to each industry ters. Given the generally weak institutional structures, and estimated the spatial distribution of these subsidies the means of stimulating the economic base of the for nineteen states in Nigeria. They found that as a result intermediate cities are even less apparent than is typi- of the spatial distribution of economic activities 89 per- cally the case in higher-income countries. cent of total net subsidies granted to industries benefited Lagos. The magnitude of these industrial incentives overwhelmed those proposed by a contemporaneous of- ingI Urbanina ficial report on industrial dispersal, which suggested m India and Chia that industrial decentralization be pursued to promote India and China have the two largest urban popula- social equity, to avoid congestion at the center, to avoid tions in the world, but by the end of this century they inflationary pressures on products and wages because of will be the only important countries with less than 50 urban congestion, to improve regional employment percent of their populations in urban areas. In spite of opportunities, and to foster political cohesion. Budget- their low per capita income levels they differ signifi- ary resources allocated for these explicit decentraliza- cantly from African countries in that they have extensive tion objectives could not be precisely calculated, but it and resilient institutional structures. was estimated that they were less than a tenth of the Until now there has been a strong bias in India against estimated industrial subsidies that benefited Lagos. the urban sector. In few other countries could one find a Concurrently the agricultural sector, which has great more firmly held view that activities to improve urban potential, performed poorly as a consequence of in- areas are by their nature antirural. Recently, however, appropriate government policies, particularly unfavor- there have been indications that this view of urbaniza- able domestic terms of trade. Agricultural exports de- tion as an urban-rural zero-sum game has been giving clined substantially, and the output of food crops, except way to greater understanding of the need for well- rice, became inadequate; food imports increased tenfold, structured urban policies. The results of the 1981 census from $200 million in 1973 to $2 billion in 1980. The are contributing to this shift in opinion. Contrary to instability of oil revenues has accentuated the difficul- expectations, the total population growth rate did not ties of managing the economy and developing urban decline during 1971-81; in fact, it increased by 24.75 policies., percent. The rate of urbanization continues to acceler- The case of Nigeria well illustrates the dominance of ate. That rate grew at 2.58 percent a year during 1951- national economic policies over the direction and loca- 61, at 3.79 percent during 1961-71, and at 4.60 percent tion of urbanization in Africa. Remarkably, however, the during 1971-81. The urban sector is expected to have concentration of new industries in a few urban centers is 300 million people in the year 2000. At the same time, making national economic planners, who are tradi- more than 50 percent of GNP is now produced in urban tionally indifferent to urban policy, more aware of the areas, even though close to 70 percent of the labor force crucial role of effective management. Even in this period remains in agriculture. of economic slowdown and scarce public resources, Given the federal structure of India and the size of its there is greater support for effective urban management states, problems with urban policies are discussed more than there was a decade ago. Sporadic breakdowns in effectively at the state level. In the three most urbanized urban public transport, which prevent workers from states of Maharashtra (Bombay), Gujarat (Ahmadabad), reaching manufacturing plants in the capitals of smaller and Tamil Nadu (Madras), as well as in West Bengal African countries, also have had a clear demonstration (Calcutta), the management of large cities is being im- effect. proved. The greatest progress has been in the acceptance Regarding the future of urbanization policies in Sub- that quality standards in the residential sector have to be Saharan Africa, the priority should be on laying solid tailored to levels of household income. But the strong Urban Development Policies in Developing Countries 69 Indian preference for regulations and for control of the ment of cities. Given present income levels, municipali- entire economy over the use of incentives to achieve ties provide high-quality urban services in garbage col- desired social objectives is still felt. Bombay's costly plan lection, street cleaning and maintenance, water supply, to freeze city growth and to shift development to New and waste disposal. Three areas which require particular Bombay is only partially helped by spontaneous sub- attention in the largest cities are public transport and urbanization. In every state the desire to relocate indus- traffic management (with the growth of motorization), try in small towns and to stop its spontaneous growth environmental pollution, and the upgrading and expan- in larger urban centers contributes to expanding the sion of the housing stock. The size of the country, the underground economy that is stimulated by conflicts level of development of the transport system, and the between licensing and the needs of business firms. As in government controls over trade and migration all con- most countries, India's coordination of urban policies is tribute to the substantial regional income disparities complicated by the split between the secretaries of com- that still remain among the twenty-six provinces and merce and industry, who regulate industrial incentives autonomous regions. and location, and the secretaries of public works and housing, who better understand urbanization. Serious Decentralization Policies and Regional problems must be solved regarding the internal manage- Development in Korea and Malaysia ment of cities in India, with special concentration on the land use legislation of 1976, the long-standing rent Two interesting urbanization policy efforts are being control in the housing sector, and the improvement of made in Korea and Malaysia. Since Korea has one of the the fiscal strength of the cities. highest population densities in the world, and only 20 Urbanization policies in China differ considerably percent of its land is usable, it has developed one of the from those in India. By substituting detailed regulations most comprehensive national land-use planning policies for market mechanisms, China has so far bypassed the anywhere, with the dual aim of controlling the rapid problem of balancing explicit decentralization incen- growth of Seoul and reducing regional economic dis- tives against the implicit incentives of trade and growth parities. Those policies attempt to redress the differ- policies. In addition, China appears to have succeeded ences in employment opportunities and in public and more than most developing countries with similar in- private services between Seoul and other regions that come levels in improving the internal management of its fuel rapid growth in the capital region. Under the land- cities. use policies, licensing for manufacturing firms in Seoul According to the 1982 census, China's population has was established, regional industrial estates were de- passed the 1 billion mark.6 The annual rate of increase is veloped, and firms already in Seoul and considered foot- 2.1 percent, and total population has increased by 313.6 loose were ordered out of the city. Educational dispari- million since 1964. The population in cities and towns ties were tackled systematically by regulating school grew at an annual rate of 2.7 percent, implying that registration in Seoul, by increasing the budgets and 206.6 million people are now living in cities and towns, enrollment quotas of provincial universities, and by re- or 20.6 percent of the total population, compared with gionalizing educational budgets. The extremely rapid 18.4 percent in 1964. Since the first plan (1953-57), the growth of the past two decades and the dispersion of policy has been to suppress the growth of large cities and heavy industrial investments around the country con- to relocate industry from the coast to the interior. Strict tributed to an equalization of regional opportunities, but control over migration through work permits and con- the rapid rise in the level of urbanization caused by trol over housing, as well as forced out-migration of export-oriented growth policies and industrialization young workers, have prevented the growth of cities, ensured that the largest urban centers, especially Seoul particularly the largest ones. For instance, the popula- and Pusan, would continue to grow. The inevitable tion in Shanghai province grew only from 10.6 million geographic decentralization of heavy industries for to 11.8 million people between 1960 and 1982. The purely physical reasons did not compensate for the fact five-year plan for 1980-85, which projected the expan- that 80 percent of manufacturing employment is gener- sion of both internal and external trade, is likely to have ated by small and medium-size firms located outside speeded up the growth of the urban sector, but there planned industrial estates, as is the case in most coun- remains a strong objection to the growth of the largest tries. The unanswered questions in Korea at present are cities. the extent to which concentration would have been While the normal growth of cities through migration worse in the absence of the policies, and the costs and and natural increase has been controlled, considerable benefits of control of industrial location. efforts have been made to improve the internal manage- Some of the most extensive efforts in Asia at develop- 70 Bertrand Renaud ing lagging regions are found in Malaysia, because of its requires the strategy. Each national agency develops its federal structure. Analytically, the most influential proj- own programs at its own pace and on its own terms. ect is the Muda River irrigation project in the northwest section of the peninsula near Thailand, which was roblems of Large Cities started in 1967 and has been the object of extensive in latin America studies since 1972. A major contribution of these studies has been the identification and measurement of the The level of urbanization in Latin America is high direct and indirect effects of large agricultural invest- compared with that of other developing regions; it ments. In addition, they show how the effects are distrib- ranges from more than 80 percent in the southern zone uted among the people of the region and examine the (Argentina, Chile, and Uruguay) to less than 25 percent extent to which project benefits leak outside the region. in Haiti. Urbanization in Latin America has been faster, The work of Bell, Hazell, and Slade (1982) is one of the more concentrated, and on a larger scale than in Euro- few which give a detailed perspective of rural-urban pean countries. Latin America's experience is similar to interactions. The authors found that, given regional that of other developing areas, but because urbanization land tenure patterns and the social structure, the in- took place earlier in Latin America than in other re- direct effects of the project were large. In particular, gions, urban institutions and policies are already well about 80 percent of additional value added was gener- developed. In spite of the spontaneous shift of growth to ated in the region for each dollar of value added directly intermediate cities-annual growth rates for which are generated by the project. This indirect income accrued now equal to or greater than those of the largest cities- mostly to nonagricultural households in the region that the population problems of the supercities still domi- worked at nonfarm enterprises. nate. Venezuela's industrial deconcentration policies As the project matured, the net capital outflow to illustrate the explicit urbanization policies in the region other regions increased significantly. Seven years after (see chapter 9, by Reio. In Brazil environmental regula- the start of the project the annual net capital outflow was tion in Sao Paulo has had a significant but selective estimated at 72 percent of regional household savings. impact on suburban relocation of manufacturing (see The leakage was attributed to the lack of investment Thomas 1981). Henderson, in chapter 7, reports on the opportunities in the region compared with altematives impact of labor market structure on industrial growth elsewhere in Malaysia. The relative isolation of the re- and location in Brazil. A World Bank research project on gion and the high cost of transport limited the growth of spontaneous industrial decentralization in the state of an industrial base. Although large direct benefits to the Sao Paulo shows that, except for suburban relocation region were provided, the results showed that self- across the rings of the metropolitan region, only limited sustaining growth was often more limited than one firm relocation takes place in intermediate cities (Hamer might expect. The analysis showed the importance of 1985). In addition, the local structure of the labor mar- induced household demand in generating downstream ket is an important factor in the emergence of new firms. benefits and the crucial role played by the sectors that Economies of scale within an industry were found to be produce nontradable goods and services. It was found more important than city size in influencing firm loca- that assistance to these sectors in the form of credit, tion. These findings cast a new light on past work by skilled personnel, and access to markets might limit net regional economists, which focused on subsidies and tax capital outflows. holidays. They also indicate that in future work the gap Another recent review of Malaysian regional policies between economic analysis and actual firm behavior and examined the cumulative impact of major programs in decisionmaking should be narrowed. the northeastem states of Kelantan and Trengganu, In Mexico efforts are being made to shape growth where economic development lags behind the rest of the patterns within the capital region through a combina- peninsula and the per capita regional GDP is 56 percent tion of transport investment, provision of infrastructure below the national average. In addition to the truly services, and regulation of investment. The dominant regional and location-specific disadvantages of the role of the capital region will not change; the priority is states, a major finding of this review was that inadequate to make the region more efficient. coordination of programs and projects implemented in the region has limited their impact. This finding for Malaysia illustrates the major problem of urbanization Conclusion policies: often there is no regional plan or any formally stated development strategy because there is no political Dominant considerations in the coming decade in the or administrative unit coterminous with the area that formulation of urbanization policies will be the general Urban Developmnent Policies in Developing Countries 71 economic slowdown, the scarcity of public resources, Notes and the continuing high pace of urban growth in spite of relative declines in demographic growth rates. Compre- 1. U.N. urban population projections are derived from the hensive national urbanization policies dominated by a estimation of urban logistic curves. See United Nations (1977). strong preoccupation with population distribution and . 2. See the recent cross-section and time-series estimation with improvement of the urban system are not likely to of an urban logistic curve by Mills and Becker (1982). be effective and should be de-emphasized. They are not 3. This profile of African urban policy reflects useful discus- realistic on several counts. First, they typically assume a sions with George Beier of the World Bank and draws from complex structure of simultaneous programs which some of his recent unpublished work. government agencies are unable to execute and which 4. See World Bank (1981), table 4.1, page 41. 5. The development of the new federal capital in Abija and simply require too much program coordination and the revision of the constitution to increase the number of execution. Second, as has happened in several advanced states are also likely to complicate the execution of urban economies, especially those with a federal structure, the policies. desire to correct the unintended effects of national eco- 6. The official figure of 1,031,882,511, as of July 1, 1982, nomic policies sets spatial policies on a collision course includes Taiwan, Hong Kong, and Macao. The mainland with other policies and policymakers. If urban develop- population is 1,008,175,288, and the average annual mainland ment ministries attempt to displace finance ministries, increase has been 17.4 million. Shanghai province has 11.8 their efforts will be futile, especially if their policies rest million, Beijing has 9.2 million, and Tianjin has 7.7 million. on weak analytical foundations. These three largest cities represent only 12 percent of the On the positive side, the past decade has seen progress urban population. on two fronts. First, analyses of urbanization policies have improved noticeably. Second, policymakers have Biblioraph clearly become aware of the essential role of the urban B grPhY economy in the national economy. Cities contain the Bell, Clive, Peter Hazell, and Roger Slade. 1982. Project Eval- leading economic sectors, they are the incubators for uation in Regional Perspective: A Study of an Irrigation many innovations, and they provide a network for the Project in NorthwestMalaysia. Baltimore, Md.: Johns Hop- flow of goods and services within a nation. As the exam- kins University Press. ple of India shows, even when a nation is still over- Bertrand, Trent, and James Robertson. 1978. "An Analysis of whelmingly rural, with 70 percent of its labor in agricul- Industrial Incentives and Location in Nigeria." Western ture, more than half of its GNP is produced in cities. Africa Regional Office, Projects Department, Industrial De- Policymakers are now well aware of the crucial need for D.C. Processed. efficiently run cities.D..Poesd Although studies such as that of Kelley and William- Hamer, Andrew Marshall. 1985. Decentralized Urban Develop- son(chaptuder 3 suhenotyet an adeqatofeey basisWilliam- ment and Industrial Location Behavior in Sdo Paulo: A son (chapter 3) are not yet an adequate basis for the Synthesis of Research Issues and Conclusions. World Bank formulation of population distribution policies, they are Staff Working Paper 732. Washington, D.C. beginning to map out the links between economic Henderson, J. Vernon. 1977. Economic Theory and the Cities. growth strategies and urbanization. Such analyses im- New York: Academic Press. prove the quality of predictions of future urban change Kelley, Allen C., and Jeffrey Williamson. 1984. What Drives and thereby facilitate the task of the public and private Third World City Growth? A Dynamic General Equilibrium institutions responsible for urban affairs. Other recent Approach. Princeton, N.J.: Princeton University Press. work also shows why urban strategies will vary accord- Lee, Kyu Sik. 1981. "Intra-Urban Location of Manufacturing ing to levels of development and urbanization. Employment in Colombia." Journal of Urban Economics, The management of urban growth should continue to vol. 9, no. 2 (March), pp. 222-41. define the policy context in developing countries. The Linn, Johannes F. 1983. Cities in the Developing World: Poli- two priorities for the decade should be better internal cies for Their Equitable and Efficient Growth. New York: management of cities by local governments and im- Oxford University Press. provements in cities' resource bases and mobilization. Mills, Edwin, and Charles Becker. 1982. "Urbanization and Urban policies should incorporate more realistic sector- Economic Development." Princeton University, Princeton, al policies regarding housing, transport, and the provi- N.J. Processed. sion of utilities. They should concentrate every effort on Renaud, Bertrand. 1981. National Urbanization Policies in seemingly unglamorous but fundamental managerial Developing Countries. New York: Oxford University Press. activities at the local level such as land registration, land - . 1984. "Structural Changes in Advanced Economies cadastres, and mechanisms for cost recovery. and Their Impact on Cities in the 1980s." In Robert D. Ebel, 72 Bertrand Renaud ed., Research in Development and Public Finance, vol. 4. tries. Vol. 10, pt. 2, in D. R. Diamond and J. McLoughlin, Greenwich, Conn.: JAI Press. eds., Progress in Planning. Elmsford, N.J.: Pergamon. Richardson, Harry W. 1977. City Size and National Spatial Townroe, Peter M., and Vinod Thomas. 1983. "Pollution Con- Strategies in Developing Countries. World Bank Staff Work- trol Policies and the Location of Industry." In T. O'Riordan ing Paper 252. Washington, D.C. and R. K. Turner, eds., Progress in Resource Management Ruane, Frances. 1981. "On Modeling the Influence of Sectoral and Environmental Planning, vol. 4. John Wiley and Sons. Policies on the Spatial Concentration of Industrial Activi- United Nations. Population Division. 1977.Manual8. Methods ties." World Bank, Washington, D.C. Processed. for Projections of Urban and Rural Population. Population Thomas, Vinod. 1981. Pollution Control in Sdo Paulo, Brazil: Studies no. 55. New York. Costs, Benefits, and Effects on Industrial Location. World World Bank. 1979. World Development Report 1979. New Bank Staff Working Paper 501. Washington, D.C. York: Oxford University Press. Tolley, George, Philip E. Graves, and John L. Gardner, eds. . 1981. World Development Report 1981. New York: 1979. Urban Growth Policy in a Market Economy. New Oxford University Press. York: Academic Press. . 1982. World Development Report 1982. New York: Townroe, Peter M. 1979. Employment Decentralization: Pol- Oxford University Press. icy Instruments for Large Cities in Less Developed Coun- 6 Urban Policy in Centralized Economies: China William L. Parish T he proper shape of Chinese cities has been a matter given the tensions with the United States in the 1950s, of intense policy discussion since 1949. There have large coastal cities were seen as vulnerable to military been some constants: since the early 1950s government attack and as improper bases for vital industries. Start- leaders have in principle consistently promoted small ing in the 1950s investment funds were siphoned from over large urban centers, tried to narrow the income gap coastal cities and reinvested in the interior. Factories between city and village, and sought to provide a secure and even some universities were dispersed from the east floor of basic needs-health, education, housing, and coast and reestablished further inland. New railroad essential food supplies-for everyone. lines were built to serve interior cities and to provide an There have also been significant changes. Restriction effectively linked interior urban network (Kirkby 1985; of the growth of cities of any size was emphasized more Lardy 1978; Leung 1980). in 1962-76 than before or since. Production rather than Also, from the late 1950s through 1976, Chinese plan- consumption was emphasized more before 1978. And ners tried to limit the growth of cities of any size. In the housing, which is defined as a consumption good, was mid-1950s Chinese cities continued to be plagued by downplayed before 1978. persistent unemployment, and national planners in- The task of this chapter is to trace the constants and creasingly came to feel that providing schools, parks, fluctuations in Chinese urban policy since 1949 and to housing, and other amenities for new urban residents assess the net impact of these policies on city size and consumed scarce resources that could be better invested stability, productive urban employment, and adequate in industry and long-term economic growth. By 1958 a urban services. tough new migration law was passed that forbade people to enter cities unless they had a residence permit. The permit could be obtained only by persons who had se- City Size and Stability cured a state job, and even then the employee's spouse and children could not accompany the new employee Post-1949 Chinese policy called for restricting the but had to remain in the countryside. rate of urban growth in general, but policymakers were A number of measures helped make this antimigra- concerned particularly with the growth of such large tion law effective. Industries and work units came under coastal cities as Shanghai. The coastal cities, many of government control and were prohibited from bidding which blossomed after contact with foreign powers in freely for labor. All new positions had to be approved and the nineteenth century, epitomized to many Chinese allocated by the government labor bureau. Also, the what was wrong with the old social order. As centers of structure of cities became increasingly cellular, with sin and decadent bourgeois habits, they violated the new each local cell or neighborhood centered around a police spartan socialist ethic. Because they concentrated station. The system remains largely the same today; wealth on the coast, those cities also fostered a regional each local police station registers every household inequality that the new leaders wanted to alleviate. And, allowed to live in the neighborhood. Periodically police 73 74 William L. Parish and local neighborhood officers conduct residency been many reasons for peasants to want to desert village checks to root out those who lack a valid residence for urban life. If that desire was to be curtailed, it had to permit. be done through administrative measures which denied Finally, the system of rationed goods provided re- peasants the benefits of urban life. straints against urban residence by anyone missed by the Chinese statistics on urban growth were previously police. Until recently, Chinese cities had one of the clouded by changes in the definitions of cities-in the longest lists of rationed goods in the world, typically inclusion or exclusion of farmers living in or adjacent to including such basic necessities as grain, meat, cooking towns, and in the size of towns counted as urban. We oil, bean curd, sugar, laundry soap, cotton cloth, coal, have made some progress in sorting out these changes, kindling, toilet paper, bicycles, sewing machines, major however, and the general trends are now clear (see Chan pieces of furniture, and, occasionally, other items that and Xu 1985; Kirkby 1985). In the 1950s city growth was were temporarily in short supply. With strict restraints very rapid, nearly 8 percent a year. In the 1960s and early on private markets, black market prices were so high as 1970s, when strict migration controls were in effect, to make it extremely difficult for a person to live in the urban growth slowed to a mere 2 percent a year. But city without proper registration for any length of time. since 1977, with the relaxation of urban controls and the Without these kinds of extreme administrative mea- promotion of small towns, urban growth has accelerated sures, cities would likely have grown much more to over 5 percent a year (Chan and Xu 1985; State rapidly, given the large gap in incomes and services Statistical Bureau 1984). between city and countryside. Despite pro-peasant ver- China's simple urban growth rates, then, have been biage, the gap between urban and rural incomes re- distinctive only in the 1960s and early 1970s. Both mained considerable throughout the first three decades before and since, China has been similar to other de- of socialist rule. The new leaders raised the prices paid to veloping societies. Table 6-1 captures part of this com- farmers, and rural taxes declined, but at the same time parison: during the past two decades, which include rural resources were siphoned off through the high both pre-1977 and post-1977 trends, cities in China prices farmers had to pay for industrial goods (Xue 1981, grew more slowly than in most developing countries, pp. 177-85). The government gave little in retum, but the post-1977 return to growth rates of over 5 spending only a little over 10 percent of its total invest- percent brought China's urban growth rates back into ment on agriculture ("Commentator" 1979). line with those of most low-income developing coun- The results of these policies are obscured by the hid- tries. den consumption urban residents enjoy over rural resi- China has been more consistently distinctive in the dents in subsidized housing, food, medical benefits, and distribution of urban growth among cities of different other social services, all of which have increased over the sizes. Much like India, which also has a large internal years. One set of national account statistics, however, market and a long history of cities scattered throughout shows that per capita consumption for the nonagricul- the country as much for administrative as for trade tural population compared with the agricultural popula- purposes, China began the modern era with several large tion continued to climb through the 1970s and declined cities rather than a single primate city that dominated only with new agricultural policies in the 1980s (State the urban landscape.2 Even after Western contact, cities Statistical Bureau 1985b, p. 552): of different sizes were spread relatively evenly across the national landscape, with only a slight tilt toward the Consumption ratio, nonagricultural large coastal cities favored by foreign traders (Berry and to agricultural population Horton 1970, pp. 67-75; Skinner 1977). 1952 2.4:1 The policies of removing even that small tilt toward 1965 2.4:1 large coastal cities was in many ways successful. The 1975 2.6:1 largest city, Shanghai, with a population of 6 million in 1978 2.9:1 1949, accounted for only 11 percent of the total urban 1984 2.2:1 population in China. In subsequent years, Shanghai grew slowly, and its share of total urban population Thus, despite its origins in a peasant revolution, declined steadily until it accounted for only 5 percent of China seems to have suffered some of the same problems total urban population in the 1960s and 3 percent in the of urban bias found in many other developing societies. 1980s (table 6-1). Even in 1984, when conditions had improved for the This pattern was repeated, by and large, for other rural population, the urban-rural consumption gap re- large cities. A comparison of 1953 and 1983 census mained as large as or larger than that in India, Pakistan, results shows how smaller cities tended to catch up with the Philippines, and Thailand (Lipton 1977, p. 430).' larger cities (State Statistical Bureau 1983; Ullman Throughout the past three decades, then, there have 1961):3 Urban Policy in Centralized Economies: China 75 Table 6-1. Urbanization in Selected Countries Urban population Percentage of urban population As percent- age of total Average In cities In largest population annual of over city growth rate, 500,000, Country or group 1960 1980 1970-80 1980 1960 1980 China' 18 21 2.7 53 5 3 India 18 22 3.3 39 7 6 Pakistan 22 28 4.3 51 20 21 Indonesia 15 20 4.0 50 20 23 Low-income countries' 11 19 5.4 40 25 28 Middle-income countries' 33 45 4.0 48 28 29 a. 1964 and 1982. b. Averages are weighted by population. The low-income average excludes China and India. Sources: China, State Statistical Bureau (1985a), pp. 54, 58, 90, 549; World Bank (1982), table 20. Annual growth, urban in official statistics.' All of this helps to explain 1953-83 Number of 1953 population (percent) cities why the proportion of China's urban population in 2,000,000 and over 1.8 4 places with more than 500,000 persons is so high in 1,000,000-1,999,999 2.1 5 relation to that in other developing countries (table 6-1). 500,000-999,999 2.5 15 The figure is in part a definitional fluke stemming from 200,000-499,999 3.1 29 the exclusion of many small places, but it also correctly 100,000-199,999 4.5 42 suggests that it was mostly the middle-level cities that 50,000-99,999 4.8 41 functioned as centers of administration and of state- In line with these growth rates, the overall rank size owned industry which prospered in the middle years of distribution of cities began to more closely approximate the new Chinese social order. the even spread of cities of different sizes that had char- Officially, the current policy is to promote the growth acterized China before Western contact in the of the smaller rural market towns which were discrimi- nineteenth century. And, as the policymakers intended, nated against previously, but there is a continuing aca- cities in the interior grew more than cities along the demic debate within China about whether this is the coast (Goldstein 1985, p. 35; Kirkby 1985, ch. 5). wisest course. Some argue that a poor country like We should not overstate the trends, however, nor China cannot afford to support inefficient small towns assume that the issue of China's future direction is (see Kirkby 1985, pp. 23042). And even if official policy settled. The shifts to the interior and toward smaller continues to promote small towns in principle, the un- cities were modest and there were important excep- usually high costs of raw materials, energy, and trans- tions-for example, the 3.5 percent annual growth rate port faced by producers in these places make their viabil- of the second largest city, Beijing.4 ity uncertain. All this may account for the high business In addition, there were many restraints on the growth failure rate reported in some sources (Fei 1984). Much of the smallest places, those with less than 20,000 as in other developing societies, failure to pay attention population. Despite protestations of support for small to-or an inability to deal with-the many indirect industry in rural towns, central planners tended to favor policy consequences of taxation, public investment, and larger industry. And the hostility to private marketing controlled prices may be what distorts well-intentioned meant that much of the informal activity that provided but eventually unsuccessful programs of small-town de- the base for small market towns was lost. Chinese schol- velopment (Linn 1983). ars now complain that many small market towns stag- If the small-town policy is unsuccessful, large cities nated (see, for example, Fei 1984). In official statistics, could be flooded with new migrants. Already, with the most of these small places even lost their urban clas- return to family-based farming, a third of the agricultu- sification after 1953, making it nearly impossible to ral labor force has abandoned simple grain production. trace their precise development in the intervening years. In addition, the reduction of the number of goods It is only in the past few years that these places have been rationed in cities and the opening of urban free markets allowed to reinvigorate their private markets, to open have decreased the number of administrative tools for more industry on their own, and to be recertified as keeping peasants out of cities. Indeed, a few cities have 76 William L. Parish already begun to issue temporary residence permits to where they could be usefully employed. In 1962 alone 20 peasants (Solinger 1985). These permits continue to million people returned to the countryside (Hu 1982). deny peasants the many subsidized rations, housing, and A second factor was the increasing employment of other services offered to normal urban residents, but women. Though this change is not immediately obvious they do attempt to regularize some of the inevitable in the table above, female employment apparently began movement into cities. to increase in the late 1950s and became almost univer- China, then, has followed a distinctive course in sal by the late 1960s. By the 1980s the vast majority of urban development, but its course has been neither a women in their twenties and thirties worked outside the simple one nor one which is certain to remain constant home-not in part-time jobs, as in many other societies, in the future. This is a time of transition in which there but in full-time jobs that kept them busy eight hours a is likely to be frequent fine-tuning to try to meet evolv- day, five and a half to six days a week. ing and sometimes unanticipated economic conditions. The 1982 census results for working-age men (16-59) and women (16-54) in cities (mostly over 100,000 population) and towns (mostly under 100,000 popula- Useful and Productive Employment tion) suggest the pattern. The employed population of both males and females was extremely high (State Sta- China's new socialist leaders wanted able-bodied tistical Bureau 1985a, tables 40, 58): members of society to have secure, productive jobs. A worker's future career was no longer to be endangered Employed Employed males females by the whim of fickle private employers but rather was to (percent) (percent) be guaranteed by the state. No longer was labor to be Cities 94 84 wasted on bourgeois consumption; it was to be invested Towns 92 78 in productive activities that led to rapid growth of indus- try and the provision of basic staples for everyone re- A third factor, which is related to increased female gardless of income. The pursuit of these goals has been employment, is the sharp drop in the urban total fertility successful, and the result is a pattern of urban employ- rate, from a high of 6.2 children in 1963 to alow of 1.1 in ment that is very different from that found elsewhere. 1980 (Coale 1984, p. 59). By the 1980s urban fertility was Nevertheless, some of the problems of unemployment extremely low, and there were far fewer children for each and poor services encountered elsewhere have also family to support. Women were freed for full-time labor, emerged in China. and the number of unemployed dependents declined. Finally, the most recent rise in employment is tied to Secure Employment efforts to create new employment opportunities (see "New Solutions," below). Urban employment has increased sharply over the While employment increased, associated security and past three decades. This is seen imperfectly but dis- fringe benefits also increased steadily for the majority of tinctly in surveys of families of workers employed in the urban population. By 1957 the state (including cen- state and urban collective enterprises. Judging from the tral, provincial, city, and county authorities) had begun inverse of the number of dependents in these families, to take over most major industries. In that year the state the employed population has almost doubled since the already employed 68 percent of the nonagricultural 1950s (State Statistical Bureau 1985b, p. 561):5 work force, and by 1982 that proportion had risen to 77 percent. Also during this period the private work force of Percent of nonagricultural artisans, peddlers, and other individual workers declined population employed steadily, and by 1982 it constituted only 1 percent of the 1964 29 nonagricultural work force.7 Those who joined state- 1978 49 owned work units enjoyed many benefits, such as health 1984 58 insurance, which paid virtually all their medical ex- penses and half the expenses for their dependents as Several factors contributed to this trend. One was the well. A pension system provided an average 70 percent policy of excluding peasants who might have come to (now 75 percent) of preretirement pay. A hardship allow- seek nonexistent jobs and of purging cities of those ance took care of those whose family dropped below a without proper employment. For example, in 1957 and certain poverty floor. There was disability insurance, 1962 there were drives to send peasants and "vagrants" lengthy sick leave, and many supplements that took care back to their former home villages or to state farms of special clothing and transport needs. Finally, it was Urban Policy in Centralized Economies: China 77 virtually impossible for anyone to be laid off from jobs in ern formal employment and traditional informal em- the state sector, and life for workers was far more secure ployment seems much more muted in China than else- than it had been in the past. These conditions still obtain where. today. Workers in a similar sector in other developing societies are sometimes referred to as the aristocrats of Serices the labor force. In China these aristocrats are more numerous and probably enjoy more wide-ranging ben- The distinction between urban work in China and efits than in most other developing societies. elsewhere is also seen in the types of job pursued. In many developing societies menial service jobs such as Informal Sector washing cars, shining shoes, peddling trinkets, and working as household help have been saturated by In spite of these advances, not all workers made it into peasants in search of nonexistent jobs in the cities. the well-paid, secure state sector. Another sector of These are exactly the kinds of bourgeois nonproductive collective enterprises, run by neighborhoods, towns, activity that Chinese leaders set out to eliminate in the and even larger state enterprises, has persisted to a 1950s. The change in the composition of the labor force degree that seems remarkable in a socialist state. These since then shows how successful they have been. collective enterprises tend to be smaller labor-intensive Today in developing societies the distribution of the units and to have lower wages, less security, and fewer nonagricultural labor force tends to mimic that of de- benefits than the average state enterprise. The contrast veloped economies such as the United States (see table between the two sectors is reminiscent of the distinction 6-2). The United States has come to have a large propor- between formal and informal workers in other develop- tion of its labor force in retail trade, restaurants, hotels, ing countries. government, education, and other services late in its Current statistics show that the average income for economic development, whereas the average developing workers in the collective sector in China is only 78 society, with a large sector of informal service workers percent of that of workers in the primary, state sector and a large government bureaucracy, has tended to (State Statistical Bureau 1985b, p. 558). The common leapfrog the gradual evolutionary process. No matter absence of pension, medical, and disability benefits for what the maturation process, however, most developed collective workers increases the gap. Also, in ways analo- economies today, like the United States, have over half gous to the situation in other developing countries, the their labor forces in information and service activities. secondary, collective sector attracts a disproportionate Socialist societies, in contrast, have tended to empha- share of women; some are older women with less educa- size "productive" work in manufacturing, mining, and tion, but some are younger. In 1984 women were only 32 construction and have downplayed such "consumption" percent of the state work force but constituted 48 per- activities as finance, retail trade, and restaurants. Even cent of the collective work force (State Statistical government has shrunk as a proportion of the labor Bureau 1985b, pp. 230-32). force, which is surprising given the increased adminis- In other ways, however, the formal-informal distinc- trative needs of centrally planned economies. China has tion in China is unlike that in other developing coun- rapidly adopted the socialist program. By the late 1970s tries. In China the secondary sector is smaller. It con- China's labor force had been remodeled into the produc- tains only one-fourth of the nonagricultural labor force, tive mold that China's leaders desired. By 1982 over half as compared with one-half in many other developing its nonagricultural labor force was in manufacturing societies.8 In China workers in the secondary sector are and mining, while only about a third was in finance, not much poorer than the primary labor force. They retail trade, government, and other service activities. make four-fifths as much as those in the state sector, This emphasis on production instead of bourgeois whereas in other countries informal workers make less consumption had substantial costs. Because of the pre- than one-half the wage of workers in the formal sector. 1978 emphasis on heavy producer industries rather than Further, the age, sex, and educational differences be- light consumer industries, supplies of consumer goods tween the two sectors in China seem modest in compari- remained minimal. Although production of radios, son to those in other countries (Whyte and Parish 1984, watches, vacuum bottles, and bicycles increased greatly, p. 32). The most striking characteristic, however, is that many consumer goods such as televisions, clothing, employment in the second sector remains relatively food, and books remained severely restricted in both secure in China. Employees are seldom laid off, just as number and variety. As demand outpaced supply, many they are not laid off in the state sector. Thus, in spite of of these goods, including bicycles, some foods, and the certain similarities, the degree of dualism between mod- better watches, had to be rationed by cumbersome pro- 78 William L. Parish Table 6-2. Nonagricultural Labor Force, by Economic Activity and Type of Economy (percent) Market economies Socialist economies United Developing Eastern Economic activity States economiesa China Europeb Manufacturing and mining 28 27 52 46 Construction 5 8 8 10 Transport and public utilities 8 9 8 11 Finance 5 2 1 1 Trade, restaurants, and hotels 21 21 11 10 Government and other services 33 33 21 22 Note: All data are for about 1970 except those for China, which are for 1982. Figures may not add to totals because of rounding. a. Twelve societies with 1970 per capita GNP of less than $1,200, ranging from the poorest, India, to the richest, Argentina. b. German Democratic Republic, Hungary, Poland, and Romania. Sources: China, State Statistical Bureau (1985a), table 54; United States, U.S. Bureau of the Census (1976), table 597; other countries, United Nations (1973), table 10, and United Nations (1974), table 40. cedures. Any increase in unrationed goods, or a change unemployment. State investment went largely into in the variety of goods, often drew a crowd of potential heavy industry, which generated only 94 jobs per 1,000 buyers to a store. yuan investment, whereas a similar investment in light Consumer supplies and services were often distrib- industry could have generated roughly 257 jobs. Thus, uted through cumbersome channels. Gone were the the state was doing little to maximize new employment peddlers and artisans who had served Chinese neighbor- opportunities (Zhao 1980). By the early 1960s newly hoods. Consumers now had to go to centralized stores expanded urban secondary schools were beginning to and service centers that might be some distance from graduate more than 2 million new job seekers every their residences. And because of the sharp decline in the year. At the same time, retirement rates were low within number of restaurants, tailor shops, bicycle repair the young labor force that had flooded the growing state shops, barber shops, and so forth, consumers not only economy during the 1950s. The result was an employ- had to shop all over town to find what they needed, but ment crisis for educated youths that could be solved only they also had to wait in long lines once there. Food by sending urban youths to the countryside, and during market lines formed before and after work and especially 1966-76, 17 million youths were sent to rural areas. on Sundays, when more people had the day off. This This makeshift solution to the employment program situation was particularly difficult for the increasing proved problematic. Many youths who could not cope number of dual-career families who had no adult at with rural living conditions slipped back into the city, home during the week. where they could not legally get a job or proper urban By the end of the 1970s bad conditions were begin- rations. ning to lead to many complaints in the press, including By the start of 1979, when the program of sending the cry that families spent so much time searching for youths to the countryside began to be phased out, there supplies on their day off that they worked harder on were 5 million to 12 million unemployed youths in the Sunday than during the rest of the week. The time spent cities, or 5 to 11 percent of the nonagricultural labor scouring for basic supplies and services left little time for force.9 Thus, China was encountering a problem com- recreation or other joint family activities. The goal of mon to developing-country cities: the rapid expansion of involving people in productive work was coming into the educational system up to the middle school (high direct conflict with the goal of improving the quality of school) level produced a new group of educated youths life for families in cities. Chinese press reports suggest with high aspirations. They hoped to get the same type of that because of the difficulties with supplies and ser- good jobs that their older siblings and neighbors had vices, many people began to perceive the quality of life in obtained, but because the value of education had de- cities as deteriorating in the 1970s. creased, those jobs were not available to them. And because of China's investment policies, jobs available in Youth Unemployment larger cities were very scarce. The result was widespread frustration and disillusionment and an outbreak of petty An additional problem caused by the emphasis on crime that alarmed many parents and central leaders production and heavy industry was increasing youth (see Whyte and Parish 1984, ch. 8). Urban Policy in Centralized Economies: China 79 New Solutions over earlier years. With more colorful clothing available and more artisans and service activities in the streets, The problems with youths and with consumer goods cities began to take on a livelier appearance. and services helped bring about a reversal in several economic and educational policies after 1978. State in- vestment shifted slightly from heavy industry to light Health Care and Housing industry, which not only produced more consumer goods but also provided more jobs. Total investment, The ambivalent attitude of many of China's leaders which was roughly 36 percent of net material product in toward urban services has been noted. Leaders have 1978, was reduced to release funds for current income been in agreement that urban dwellers should be sup- and for an increase in consumption to promote the plied with such basic human needs as grain, essential growth ofjob-producing light industry. Collective enter- cotton clothing, and an initial ten years or more of prises supervised by neighborhoods and towns were en- education. It is widely accepted that these should be couraged-employment in this sector rose from 22 per- available to everyone on a relatively equal basis. There cent of the nonagricultural labor force to 26 percent. has been less agreement on social services and supplies And more people were allowed to go into business on beyond the basic needs. The volume of goods and ser- their own: the share of self-employed persons in the vices supplied and the degree of insistence on equal nonagricultural labor force rose from 1 percent in 1978 distribution of those goods and services have shifted to 3 percent in 1984 (State Statistical Bureau 1985b, with the political winds. p. 214). Peasants were permitted to come into the mid- dle of cities to sell their goods in newly reopened peasant Health Care free markets, and the variety of goods available to con- sumers increased. Some full-scale middle schools that In China, as in other socialist states, health care has were turning out graduates without job-related skills been consistently defined as a basic human need and a were closed, and more technical schools to train stu- human right, but the degree of insistence on equal dents in such skills were opened. distribution of the services, particularly to the country- The results of the changes have been a significant side, has varied. Throughout, provision of ample medi- drop in unemployment and a considerable increase in cal care has had a high priority. As a result, medical consumer goods and services. Not everyone has gotten a services have grown steadily. By the 1970s the number state job. Two-fifths of those employed over the past few of persons for each doctor, nurse, or hospital bed was years have been employed outside the state sector, many much lower in China than in other developing societies, in neighborhood collective enterprises or in individual as the table shows (n.a. signifies not available)."0 activities that they started themselves, thus receiving more variable income and fewer fringe benefits than Population per service unit they might wish (State Statistical Bureau 1984, p. 130). Low-income Middle-income Some employees ended up in temporary jobs without a Service unit China countries countries secure future. Nevertheless, by the end of 1981 outright Western doctor 2,470 9,900 4,310 unemployment had declined to only 2.4 percent of the Doctor, any type 1,172 n.a. n.a. nonagricultural labor force, and the remaining unem- Nurse 2,306 8,790 1,860 ployed were mostly those who had graduated from mid- Hospital bed 503 1,730 570 dle school during the previous year (State Statistical Bureau 1985a, table 68; Zhongguo BaikeNianjian 1984, With this wide distribution of services, the reported pp. 619-20). infant mortality rate in China dropped to only 49 per Supplies and services also increased. With more 1,000 live births-less than half that of India-and the emphasis on light industry, state stores began to carry life expectancy at birth rose to age 68, eight years more more clothing, bicycles, tape recorders, and other con- than in the average middle-income developing country sumer goods. Neighborhoods began to have more collec- and eighteen years more than in the average low-income tive restaurants, laundries, bicycle repair stalls, and developing country (see note 10 for data sources). other service centers that catered to everyday needs. Health care in cities was even better than these national Peasant stalls on designated streets increased the variety averages suggest. and freshness of eggs, fish, peanuts, vegetables, and Cities are particularly well served by this system. In other goods available to urban residents. Many items large cities provision of care is three-tiered. Large work were still rationed, and shortages continued to occur for units and many neighborhoods have a first aid station some items, but the supply was considerably improved staffed by nurses and paramedics to treat minor injuries. 80 William L. Parish People with more serious illnesses go to the local hospi- villages it was 13 percent. The causes of these differences tal that is assigned to their neighborhood or work unit. If rest heavily on food supplies and public sanitation- the hospital evaluates a problem as being more serious, conditions that are difficult to change in widely scat- it can send the patient to a city or provincial hospital for tered villages. But to some extent the differences are also specialized care. The system is not without its problems. attributable to state expenditure patterns that favor Since those employed in the state sector receive essen- urban consumption and state medical expenditures tially free medical care, the patient load is high at local (which, on a per capita basis, are almost ten times hospitals. Emergency cases with raging fevers and other greater in cities than in villages), and to a similar bias signs of trauma can be taken immediately, but others toward cities in the supply of doctors, nurses, and facili- may have to line up early in the morning to take a ties (Jamison and others 1984, pp. 11, 32, 93). In health, number for an appointment that day. Complaints arise as in overall income, it has been difficult to remove some because of the lack of choice of doctor and the few of the tilt toward cities, and China again repeats some of minutes of attention the doctor provides. Another prob- the patterns found in other developing societies. lem is that medicines are sometimes in short supply in relation to demand."1 But, overall, the system seems to have worked well in supplying basic care, either free or Housig at low cost, compared with the care available in most Housing has been subject to sharp fluctuations in developing-country cities. policy. In the early 1950s there was an attempt to clean In large cities, preventive public health activities are out major slum areas and erect new apartment buildings also well organized. The city periodically organizes, in their place. But by the mid-1950s housing was de- through the cellular neighborhood and work-unit struc- clared a consumption good, undeserving of major new tures, inoculation campaigns to ensure that everyone is investment. It was not until after 1976 that planners protected against disease. Neighborhoods and work once again chose to provide major resources for housing units also help publicize the need for sanitation and for the urban population. the eradication of pests and vermin. Cities organize the The consequences stand out sharply in statistics on daily collection of human sewage from homes and public floor space per capita. From a high of 4.5 square meters latrines as well as the daily sweeping of major streets. per capita in the early 1950s, urban floor space declined Neighborhoods help organize the sweeping of inner steadily until it reached a low of 3.6 square meters in streets, garbage collection, and the seasonal cleaning of 1978. Only with massive urban building in recent years houses. For seasonal cleanings the neighborhood may did housing reach a new per capita high of 4.6 square distribute fumigating agents and then have everyone meters in 1983 (Zhongguo Baike Nianjian 1984, p. 622; gather for a movie while fumigating is taking place. Zhou 1979).12 Neighborhood leaders may also inspect each house or By some indicators, urban housing was not all that apartment to make sure that it has been properly terrible even in the middle 1970s. The stringent limita- cleaned and leave a small colored sticker on the front tions on migration averted the growth of shanty towns door to indicate that the house has passed inspection. crammed with peasants seeking urban work. And These public sanitation activities are less thorough in according to one set of statistics, based on some of the small towns and cities, but they help make China's large more prosperous regions of south China, Chinese urban cities much cleaner than many other developing- housing was as good as or better than housing in many country cities. other cities of the developing world (Whyte and Parish The conditions listed above, when combined with 1984, p. 78). These statistics imply that Chinese urban subsidized food supplies and ample nutritional support, dwellers were no more likely than people in other de- dramatically improved urban health. The rural pattern is veloping-country cities to be crammed three to a room more mixed. Nationwide, in 1975 the average life expec- and they were somewhat more likely to have electric tancy for city dwellers was 72 years, while the figure for lighting and a kitchen, although that kitchen might people in the countryside was only 57 years. In part, the have to be shared with others in the same building. fifteen-year gap documents the extraordinarily good Nine-tenths lived in brick and concrete structures in- urban health care and prevention system rather than the stead of makeshift wood and tin shantytowns. neglect of the countryside, but it is an astonishing gap, Nevertheless, there were many serious problems with nevertheless, and one matched by few other developing housing in the 1970s. The press compained that 35 countries. The gap is reflected also in rates of nutritional percent of all urban families lived in overcrowded condi- stunting as measured by the percentage of children less tions, that 20 percent lived in slums, and that 5 to 6 than 90 percent of the average height of their age group. percent had no proper housing at all ("Summary" 1979; In cities the 1979 rate was only 3 percent, while in Zhou and Lin 1980). Not only were individual families Urban Policy in Centralied Economies: China 81 crowded, but they also often had to share toilet and by centralized city housing authorities but by work units kitchen facilities with other residents in the same build- (factories, schools, and bureaus). Nationwide, in 1983, ing. This sharing was and continues to be a frequent 58 percent of housing was owned by work units, 26 source of tension between families, with one family percent by city housing authorities, and 17 percent by complaining that the other is leaving garbage scattered private individuals (Zhongguo Baike Nianjian 1984, about in the kitchen or is monopolizing precious bath- p. 623). With the emphasis on work unit housing, richer ing and toilet facilities. Indeed, one account of crime in worker units with housing that was relatively good to Shanghai claims that such disputes between neighbors begin with provided even better new housing. Poor work continue to occupy an inordinate amount of police units built little housing, as did most private individuals attention (Zhao 1984). living in their old family homes. It was a situation of the Another difficulty with public housing, common to rich getting richer while the poor stayed poor, at least in many socialist societies, is that rents are set too low to their housing. cover yearly upkeep as well as building costs (Zhongguo Another contributing factor was low rent, which had BaikeNianjian 1981, p. 541). Rent was only 1.4 percent two unforeseen consequences: rents provided insuf- of 1984 urban family budgets, and expenditures for rent, ficient money for new building or maintenance, and electricity, water, and cooking and heating fuels totaled since rent was less than 2 percent of a family's budget, it only 4.2 percent, considerably less than the 10 percent placed no restraint on overuse. Families with more typical of other developing societies (Lluch and others space than they needed felt no compulsion to move to 1977, p. 40; State Statistical Bureau 1985b, p. 567). As a smaller quarters, and local housing authorities were consequence of the low rents and the inadequacy of state powerless to get them out in favor of larger families who budgets to make up the difference, many existing build- desperately needed more space (Liu 1984, p. 28). ings deteriorated (Zhongguo Baike Nianjian 1981, Some issues with housing remain, then. If rents con- p. 541). One press source estimated that "more than 50 tinue to be low, they will provide little income for new percent of the houses in the urban areas of China are in a building or for repair, and they will fail to ration scarce bad state of repair and need maintenance, of which more housing resources. If work units continue to build much than 10 percent are in dangerous condition" (Zhou and of the new housing, some of the old inequalities will Lin 1980). remain. And, as in the past, bureaucratic allocation of In the 1970s, there were additional complaints from most housing will limit choices in both housing and professionals and administrators forced to share the neighbors. space standards and amenities of persons in blue-collar jobs. Other complaints stemmed from inability to Conclusion choose one's house or neighbors (or to move away from contentious neighbors) when housing was allocated China's history since 1949 illustrates both the poten- through bureaucratic channels (Whyte and Parish 1984, tial and the disadvantages of urban development in a pp. 76-85). centrally planned economy. With much greater control Since 1978 some of these problems have been allevi- over economic resources than in the average market ated. Floor space has increased, particularly for profes- economy, the government was able to shift investment sionals and administrators. And in new buildings with a funds to promote the development of medium over large toilet and kitchen for each apartment there is less reason and interior over coastal cities. All of this helped reduce for friction among neighbors. regional inequalities. In addition, the control over jobs Some problems remain, however. One study in Bei- and rationed consumer supplies meant that for a time jing complains that, despite a massive building program the government was able to limit severely the growth of in that city, the percentage of families living in severely all cities, and funds that might have been spent on an overcrowded housing remained unchanged through elaborate urban infrastructure for waves of new mi- 1982 (Liu 1984). "Severely overcrowded" is defined as a grants was spent instead on rapid industrial growth. per capita living space of less than 2 square meters, no As in other centrally planned economies, the govern- room of one's own after marriage, or three generations ment has been able to put the urban population to work. in the same room. Even while average floor space in Most able-bodied women and more than half the urban Beijing increased from 4.6 to 5.7 square meters in four population have jobs. Few of those jobs are part-time or years, the percentage of severely overcrowded house- likely to be lost tomorrow; they are primarily full-time holds remained at 14 percent. jobs that promise to last for a full career. In these ways Several reasons, in addition to a growing number of Chinese cities have avoided some of the problems of marriage-age youths, contributed to this problem. One employment instability that have afflicted other develop- was the increasing tendency for housing to be built not ing-country cities. 82 William L. Parish There have been policy mistakes, however, that have care, and administration that is the envy of many other created in China some of the same problems found in societies. The government has tried to help directly by other societies. It was a mistake to downplay the role of raising prices for agricultural products. But a closer light industry and consumer services. Some developing- examination shows that before the 1980s the chimera of country cities may have too many people in informal rapid industrial growth helped shape government in- service activities, but China illustrates the problems of vestments and industrial prices in such a way that agri- the opposite extreme. With so little opportunity for light culture grew less rapidly than might have been expected. industrial growth and with bureaucratic restraints on And when it comes to state-subsidized services, includ- small, individualistic enterprises, problems of youth ing housing, health care, education, and cheap food, it is unemployment began to appear. The unemployment still the urban sector that has reaped the most benefits. problem was exacerbated by the rapid expansion of em- These characteristics that China shares with many other ployment for women. Some jobs were created during the developing societies illustrate the continuing problems 1960s and 1970s, but since those jobs were taken almost of urban and industrial bias even in a society that has as frequently by women as by men, the need to create overtly renounced this route to development. additional openings was greater than in other develop- China has begun to correct many of these difficulties. ing societies. Furthermore, the rapid expansion of sec- The government has induced a rapid rise in peasant ondary school education, as in many other developing incomes by further raising rural purchase prices and by societies, contributed to a growing corps of unemployed removing many administrative constraints on farmer's educated persons. China avoided this problem at the activities. The promotion of small market towns and university level but not at the secondary level. The un- new collective enterprises is among the measures which employed youths were particularly frustrated because have helped increase peasant incomes. As a result, some the same level of education had guaranteed good jobs individual farmers have reaped rewards unimaginable just a decade before, when education was less common. even to the highest-paid urbanite. On a national basis as This is a frustration shared with many youths in develop- well, farmers have finally begun to narrow the gap be- ing societies, but it was perhaps felt more acutely be- tween average rural and urban incomes. cause of the socialist promise of secure jobs and rapid Some of these policies still hang in the balance. development. That frustration contributed to the out- Whether small market towns will indeed prosper and break of petty crime in the 1970s and continues to fuel absorb much of the excess rural labor force remains the social alienation of some youths today. As we have uncertain. Investments in local transport, small trucks, already noted, the slighting of light industry and ser- cheap gasoline, and cheap electricity may have been vices, as well as the minimal investment in housing, also delayed too long. Without these measures, farmers are helped create a perception among many urbanites that likely to find economic activities in small towns unviable the quality of urban life in the 1970s was declining and and clamor to go to large and medium-size cities. the socialist promise was going unrealized. Already, as a result of the increasing role of peasant In brief, China's experience in virtually eliminating free markets in cities, the reduction in the number of small informal service activities illustrates the necessity rationed goods, the undertaking of new construction for these types of activities in developing cities. Hacken- that requires unskilled labor, and the general relaxation berg (1980) argues that because of high population of administrative control, the number of quasi-legal growth rates and the difficulty of centralized provision of migrants in cities is rising. Increasing foreign trade may essential urban services, the informal service sector in also create more pressure for major port cities such as poor societies is essential and should be embraced rather Shanghai to grow in ways that they have not in the past. than shunned. Although they reject the extreme version Thus, while much of the basic structure may remain in of this position, China's leaders seem to be reluctantly place, we may again see a rapid growth of large cities that moving in this direction. Smaller and more makeshift will make China more similar to other developing coun- work arrangements, organized ad hoc by neighborhoods tries. Only close attention to the details of prices, taxes, and individuals, with lower rates of pay and security, are and investment will provide either observers or Chinese now approved as a way of providing both employment planners a clue as to which way this set of development and essential urban services. forces will tilt. China's experience also illustrates the difficulty of eliminating the gap between city and countryside. In many ways China seems to have worked much harder at Notes eliminating this gap than have other societies. Through the encouragement of bootstrap operations, villagers 1. The comparative data are mostly for the 1960s, and the have been induced to help build an infrastructure of ratios for these countries range from a low of 1.1 to a high of waterworks, roads, level fields, schools, public health 2.2.WiththeadditionofChineseurbansubsidies,whichinone Urban Policy in Centralized Economies: China 83 source are said to have totaled 164 yuan per capita in 1981, the lents of the Russian feldsher, but no "barefoot" paramedics. true 1984 Chinese urban-rural ratio could approach 3:1. This The data for China are for 1979; data for developing market would make the comparison with other countries even more economies are for about 1978. striking-although public expenditures and subsidies will 11. Evidence on how minimal costs to the user lead to havesomeurbanbiasintheothercountriesaswell[seeBeijing overusage comes not only from interviews but also from Review, no. 43 (1982), p. 71. Chinese drug use studies and govemment attempts to contain 2. A primate urban structure is one in which much of the costs (see Jamison and others 1984, pp. 67, 99). total urban population is in the largest city. 12. The floor space figure excludes kitchen, lavatory, and 3. Sex ratios provide an additional indicator of the faster public corridors. growth of small places. Chinese cities (mostly over 100,000 population) have 108 males for each 100 females, only slightly higher than the ratio for rural areas of 104, which suggests that a modest number of males have entered cities. In contrast, Bibliography towns (mostly of 10,000-100,000 population) have ratios of 116, implying much more migration. In the 30-34 age group, Berry, Brian J. L., and Frank E. Horton. 1970. Geographic which includes many new male laborers moving from the Perspectives on Urban Systems. Englewood Cliffs, N.J.: countryside, the ratios are 112 for large cities, 106 for rural Prentice-Hall. areas, and 129 for towns (State Statistical Bureau 1985a, tables Chan, Kam Wing, and Xueqiang Xu. 1985. "Urban Population 35-37). Thirteen million peasants took work in towns and Growth and Urbanization in China Since 1949." China cities in the decade 1966-76. These peasants were probably Quarterly, no. 104, pp. 583-613. moving to towns even while 17 million youths were being Coale, Ansley J. 1984. Rapid Population Change in China, expelled from larger cities (see Hu 1982). 1952-1982. Committee on Population and Demography, 4. Among cities with over 100,000 population, the regres- Report 27. Washington, D.C.: National Academy Press. sion slope of city size on city rank (both logged) flattened only "Commentator." 1979. "The Aim of Socialist Production Must slightly, from - 0.95 to - 0.88, between 1953 and 1982 (raw Be Really Understood." People's Daily, October 20, pp. 1-2. data from State Statistical Bureau 1983; Ullman 1961). 5. Between the 1953 and 1982 censuses, the number of Emerson, John Philip. 1965. Nonagricultural Employment in designated urban centers was almost halved, from about 5,500 Mainland China. U.S. Bureau of the Census, International to 2,990. Many places regained their urban designation in Population Statistics Reports, Series P-90, no.21. Washing- 1983, and 10,000 places were eventually to be designated ton, D.C.: Government Printing Office. urban, in line with their new economic role (Goldstein 1985, Fei, Xiaotong. 1984.Xiao Chengzhen Da Wenti [Small Town, p. 67). These changing designations probably had only modest Big Problem]. Jiangsu Province Small Town Research effects on the growth rates reported above. Some of the reduc- Group, Compendium 1. Jiangsu People's Press. tion in centers occurred before the 1966-76 slowdown in Goldstein, Sidney. 1985. Urbanization in China: New Insights growth, and the 1978-82 upturn in growth occurred before from the 1982 Census, Papers of the East-West Population new towns were designated. Institute, no. 93. Honolulu: East-West Center. 6. The results in this table are similar to the results of the Hackenberg, Robert A. 1980. "New Patterns of Urbanization in 1982 census, which showed that 56 percent of the population Southeast Asia." Population and Development Review, vol. in cities and 53 percent of that in towns were employed (State 6, pp. 391-420. 7 State Statistical Bureau 1984 pp 45 111. By 1984 the Hu, Mengzhou. 1982. "Solution to Employment Problems." 7.~~~~~~~~~ Stt'ttsia ueu18,p.4,11 y18 h Beijing Review, no. 39, p. 23. proportions in state, collective, and private enterprises were ' R n 71, 26, and 3 (State Statistical Bureau 1985b). Jamison, Dean T., and others. 1984. China: The Health Sector. 8. The comparisons here and below are based on Mazumdar Washington, D.C.: World Bank. (1976). Kirkby, R. J. R. 1985. Urbanization in China. New York: 9. The low figure, which ignores urban youths still legally Columbia University Press. assigned to agricultural jobs in the countryside, is from Hu Lardy, Nicholas. 1978. Economic Growth and Distribution in (1982). The higher figure, which includes urban youths who China. Cambridge: Cambridge University Press. still were in the countryside but were scheduled to return Leung, C. K. 1980. China: Railway Patterns and National legally to the city in 1979-80, is implicit in Zhongguo Baike Goals. Department of Geography, University of Chicago. Nianjian (1981), pp. 627-28. There is a third and even higher Research Paper 195. Chicago. estimate that may erroneously include all urban youths ever Li, Xiannian. 1979. "The Economy of the Whole Country Is consigned to the countryside, regardless of their current resi- Seriously Out of Balance."Ming Bao (Hong Kong), June 14, dence and job status. This estimate, 20 million unemployed at 1979, p. 4, in FBIS, June 19, 1979, pp. L11-13. the start of 1979, comes from a purported speech by Li Xian- Linn, Johannes F. 1983. Cities in the Developing World. New nian to the National People's Congress (Li 1979). Yr:Ofr nvriyPes 10. World Bank (1980a); World Bank (1980b); Zhongguo York: Oxford University Press. BaikeNianjian (1980); Foreign Broadcast Information Service Lipton, Michael. 1977. Wny Poor People Stay Poor. Cam- (FBIS), April 30, 1980, p. 19. "Doctor, any type" includes bridge, Mass.: Harvard University Press. Chinese herbal doctors and secondary-school-trained equiva- Liu, Zhi. 1984. "Ten Contradictions Beijing City's Develop- 84 William L. Parish ment Faces." Shehuixue yu Shehui Diaocha, no. 1, United Nations. 1973. Demographic Yearbook, 1982. New pp. 23-32. York. Lluch, Constantino, and others. 1977. Patterns in Household _ 1974. Demographic Yearbook, 1973. New York. Demand and Saving. New York: Oxford University Press. U.S. Bureau of the Census. 1976. Statistical Abstract of the Mazumdar, Dipak. 1976. "The Urban Informal Sector." World U.S., 1976. Washington, D.C.: Government Printing Office. Development, vol. 4, pp. 655-79. : \ Whyte, Martin K., and William L. Parish. 1984. Urban Life in Skinner, G. William. 1977. The City in Late Imperial China. China. Chicago: University of Chicago Press. Stanford, Calif.: Stanford University Press. World Bank. 1980a. World Development Report 1980. New Solinger, Dorothy J. 1985. "'Temporary Residence Certificate' York: Oxford University Press. Regulations in Wuhan." China Quarterly, no. 101, pp. 98 . 1980b. World Tables: The Second Edition. Baltimore, 103. Md.: Johns Hopkins University Press. State Statistical Bureau. 1983. Ten Percent Sampling Tabula- _. 1982. World Development Report 1982. New York: tion on the 1982 Population Census. Population Census Oxford University Press. Office, State Council, and Department of Population Statis- Xue, Muqiao. 1981. China's Socialist Economy. Beijing: For- tics. Beijing: Chinese Statistical Press. eign Languages Press. 1984. Zhongguo Tongii Nianjian [Chinese Statistical Zhao, Jian. 1984. "Criminal Offenses and Social Policy." She- Annuall. Beijing: Chinese Statistical Press. hui, vol. 4, no. 5, pp. 16-18. - . 1985a.;1982 Population Census of China (Results of Zhao, Lukan. 1980. "A Probe Into Some Employment Prob- Computer Tabulation). Population Census Office, State lems." Beijing Review, no. 43, p. 20. Council, and Department of Population Statistics. Beijing: ' Chinese Statistical Press. ZhongguooBakeiNianCian 1980. [CheseEncyclopediaYear- - 1985b.Zhongguo TongjiNianjian [Chinese Statistical book.1Beijing:ChineseEncyclopediaPress. Annual]. Beijing: Chinese Statistical Press. . 1981. Beijing: Chinese Encyclopedia Press. "Summary." 1979. Summary of Guangming Daily article in . 1984. Beijing: Chinese Encyclopedia Press. Xinhua, Beijing, November 17, in FBIS, November 20, 1979, Zhou, Jin. 1979. "Housing China's 900 Million." Beijing Re- p. L5. view, no. 48, p. 18. Ullman, Morris B. 1961. Cities of Mainland China: 1953 and Zhou, Shulian and Senmu Lin. 1980. "On the Question of 1958, Series P-95, no. 59. Washington, D.C.: U.S. Bureau of Housing." People's Daily. August 5. the Census. Part III Concentration or Decentralization: Evaluation of Policies p art III is devoted to the issue of urban concentration A model is presented that incorporates both urban- _and related policy initiatives. Although attention ization and localization economies, and the results of has focused in recent years on decentralization policy, estimation are reported. The model assumes constant that is by no means the only or major direction of returns to scale with a Hicks-neutral shift factor external government intervention. This part reviews policies to to the firm. The shift factor contains a variable that foster both concentration and deconcentration and reflects localization economies, as measured by own- brings out the cost of obtaining the desired results. industry employment, and a variable that reflects ur- Henderson (chapter 7) discusses the experience with banization economies, as measured by urban popula- concentration in Brazil, and Lee (chapter 8) and Reif tion. (chapter 9) analyze decentralization policies in Co- Two estimating equations derived from the theory are lombia and Korea and in Venezuela, respectively. tested for three industries in the region. One equation directly evaluates effects on production, where value Impetus to Urban Concentration added per capita is regressed on the capital-labor ratio, urban population, own-industry labor force size, dis- Chapter 7 deals with the spatial concentration of in- tance from urban area, labor force quality, and average dustry and cities in Brazil and presents the essential firm size (which is included to test the assumption of principles of industrial location theory and their rele- constant returns to scale). The second equation re- vance to Brazil. Small urban areas in Brazil conform to gresses the capital-labor ratio on factor prices, distance theoretical expectations, but the findings suggest that to urban area, property taxes, own-industry size, labor the central government has in the past promoted the quality, and firm size, where the variable for own- concentration of heavy industry in Greater Sao Paulo industry size tests the assumption of Hicks neutrality. and Rio de Janeiro. According to the author political The author finds no evidence of significant urbanization considerations may have contributed to the past bias in economies, but he discovers strong support for localiza- favor of the major urban centers, but there was also a tion economies, which tend to fall as a sector's labor belief that such a policy would foster economic develop- force grows. Results also appear to justify the assump- ment. The policy's desirability would hinge on whether tions of constant returns to scale and Hicks neutrality. It scale economies exist in large metropolitan areas and, if is concluded that a relaxation of government- so, whether they are urbanization or localization econo- encouraged concentration in Brazil would be beneficial. mies. Urbanization economies-the benefits of city size in itself, such as the size of the total labor market and the presence of communication networks-would justify Locational Choice and Decentralization concentration. Localization economies-efficiencies in labor markets and services specific to an industry and Chapter 8 addresses employment location policies. benefits arising from specialization by firms-would Deconcentration policies can be intended to modify not. location patterns within an urban region or at the 85 86 Part 117 national level. The former involve intrametropolitan fluenced industrial location, and what has been their phenomena, the latter interregional issues. This chap- relative importance? ter, which concentrates on intrarnetropolitan issues, The hypothesis that less protected industries are investigates pattems of employment location, intro- more concemed with locational factors is supported by duces an economic apparatus to explain firm location the evidence. More generally, the model attempts to choice, and tests the importance of several policy instru- identify factors that affect the location decisions of ments. firms. Altemative explanatory factors-such as changes The investigation of employment location describes in the Venezuelan economy and the unintended spatial the spatial distribution of manufacturing, commerce, effects of trade and macroeconomic policies-arose finance, and services in the Colombian cities of Bogota when little evidence was found that government incen- and Cali. Evidence is found of industrial decentraliza- tives affected a firm's location. In particular, the study tion, although not in all sectors. A life-cycle classifica- shows that entrepreneurs attach a relatively small tion of firm development (birth, death, relocation, and weight to government financial incentives. stability) reveals the important role of births and deaths in this context. The analysis provides an introduction to contemporary location theory, including developments .b1rh in stochastic specification. The chapter lists site charac- Biblography teristics which are subject to policy manipulation and Haurin, Donald. 1983. "The Incidence and Effects of Factor reports on statistical tests of their importance for three Taxes and Subsidies upon Urban Areas." Ohio State Uni- industries. It concludes with a discussion of a conscious versity, Columbus. Processed. decentralization policy with Seoul, the capital city of the Henderson, J. Vernon. Economic Theory and the Cities. New Republic of Korea, as an example. York: Academic Press. 1977. Lee, K. S. 1982. "A Model of Intra-Urban Employment Loca- tion: An Application to Bogota, Colombia." Journal of Ur- Impact of Decentralization Policy ban Economics, vol. 12, no. 3 (November), pp. 263-79. Mills, Edwin S., and Charles Becker. 1983. "Indian City Sizes In light of its experience in the 1970s with govem- and City Growth." Princeton University, Princeton, N.J. ment intervention in the spatial distribution of manu- Processed. facturing industry, Venezuela provides an opportunity Murray, Michael P. 1982. "Here, There, Where? A Strategy for to examine the effectiveness of a decentralization policy. Evaluating Industrial Relocation Policies in Korea." Urban Chapter 9 asks whether govemment instruments, in- Development Discussion Paper 6. World Bank, Washing- cluding incentives and restrictions, have contributed to ton, D.C. the deconcentration of industry in Venezuela. This ques- Renaud, Bertrand. 1981. National Urbanization Policy in De- tion leads to more specific questions. Has deconcentra- veloping Countries. New York: Oxford University Press. tion of industry taken place? Is deconcentration a result Scott, Ian. 1982. Urban and Spatial Development in Mexico. of the policy instruments? What other factors have in- Baltimore, Md.: Johns Hopkins University Press. 7 The Analysis of Urban Concentration and Decentralization: The Case of Brazil J. Vernon Henderson T he spatial patterns of industrial and urban develop- promote concentration and decentralization is not the T ment in Brazil have received considerable attention purpose of this chapter. in recent years. Discussions have been characterized by swings in opinion and policy focus that have ranged from the view that the increasing agglomeration of ac- Urban and Regional Patterns tivities has been desirable to the view that the evolving spatial concentration and congestion have been exces- The size and growth rates of the largest metropolitan sive. The relative emphasis on these views has changed areas, in particular the Greater Sao Paulo Metropolitan over time and has varied with the overall fortunes of the Area and, to some extent, Rio de Janeiro, have been economy; concern about concentration has often dimin- matters of concern in Brazil. The annual growth rates of ished during times of economic difficulty. both urban centers have slowed: for 1970-80 Greater The differing opinions draw attention to the need for Sao Paulo had a growth rate of 4.4 percent-the same as an economic analysis of the forces behind the trends the national average growth rate for urban areas-and toward concentration and decentralization. Neither Rio's rate was considerably less, 2.6 percent. Policymak- phenomenon may be undesirable in itself, but govern- ers, however, continue to view the sizes and growth ment policies may have intended and unintended im- rates of these cities as excessive. This concern is ex- pacts on industrial location, and these impacts may pressed in a medium-size-cities program which is de- affect economic welfare. This chapter examines concen- signed to attract rural migrants into medium-size rather tration as it is influenced by government policies and than large cities. attempts to evaluate the desirability of such influences Although Greater Sao Paulo, with 12 million people in when the existence (or absence) of certain types of econ- 1980, and Rio, with 8 million, are large by international omies of scale in urbanization and industrialization is standards, they are not inordinately big metropolitan taken into account. The data base for the core of the areas for a country as large as Brazil, particularly in view paper relates to 1970, and much of the discussion there- of the falling or stagnant growth rates of the two centers. fore concems the historical evolution of policies which Furthermore, the size distribution of urban areas in essentially promoted concentration. In the 1970s dis- Brazil is similar to that of such other large decentralized tinct efforts were made to achieve decentralization, and countries as the United States in terms of the proportion some aspects of those efforts are covered in part IV. An of the total population living in big and small urban up-to-date evaluation of the net outcome of efforts to areas. Small and medium-size cities are growing at least as fast as larger cities, and there are many other Note: The funding of this research by the World Bank is gratefully flourishing metropolitan areas in Brazil in addition to acknowledged. The encouragement and support of Douglas Keare and Greater Sao Paulo and Rio. Andrew Hamer was much appreciated. Tri Pham, as well as William Dillinger and David Keen, provided able research assistance. Responsi- Although the general size distribution of cities in bility for all errors is mine. Brazil may be reasonable by world standards, the spatial 87 88 J. Vernon Henderson distribution of resources between large and small cities savings except for the investments of traditional planta- may be less than desirable. In particular, Greater Sao tion familes. The government owns almost all public Paulo's industrial composition is unusually concen- utilities, 50 percent of the iron and steel industry, and 35 trated in relation to that of other large diversified econo- percent of the chemical industry, including probably mies. Greater Sao Paulo has a great concentration of most petrochemicals. The government uses its own- heavy industry within its boundaries compared with ership to influence the location of economic activity and London, New York, Los Angeles, and Tokyo, which, in the spatial and intersectoral allocation of capital. In comparison, are oriented toward services, high technol- addition, tariffs and taxes have indirectly influenced ogy, and light manufacturing. In 1970 Greater Sao location and urban concentration. Paulo accounted for 43 percent of value added for all Brazil has a federal system. Because of changing insti- Brazil in iron and steel and fabricated metals, 71 percent tutional arrangements it is difficult to determine the in transport equipment, and 35 percent in chemicals. degree of local autonomy at any one time. In 1970 some Such concentration deserves to be explained, and its funding for urban infrastructure investments such as implications for the efficiency of spacial allocation of municipal roads, water, and sewers came from national resources need to be analyzed. earmarked taxes. Much of the funding for both capital Greater Sao Paulo and Rio are located in the South and operating expenditures, however, was from local and Southeast of Brazil, a region which comprises the property taxes and user charges, and localities deter- states of Minas Gerais, Espirito Santo, Rio de Janeiro, mined their own levels of investment and maintenance. Sao Paulo, Santa Catarina, and Rio Grande do Sul. It is (Except for the state of Sao Paulo today, state agencies the most developed region in Brazil and is larger than have taken over much of the municipal decisionmaking, Spain, France, the Federal Republic of Germany, and the but this was not so for the period of our data.) Electrical Democratic German Republic combined. Our sample hookups were generally available in the 126 urban areas, consists of the 126 urban areas in this region which had although the quality of service as indicated by brown- populations over 20,000 in 1970. These urban areas now outs and blackouts apparently varied considerably. contain about 40 million people. Municipalities controlled their own planning, zoning, Highly developed and industrialized, this region is and parks but not the educational system or police and crisscrossed by major modern highways and rails, and fire protection. its cities offer the basic range of modern utilities and Thus, in 1970 localities seemed to have some degree services. Most of the cities are located either in fertile of autonomy in determining their own physical appear- agricultural areas or in areas rich in mineral deposits. ance and structure, so that urban areas could compete The region accounts for almost all of Brazil's manufac- on an equal footing for resources. Since city councils turing, and its standard of living is high for a developing (although not mayors) were locally elected, local prefer- country, despite notable deficiencies.' The principal ences for urban structure may have been represented shortcoming is in education-in the quality of the cur- and to some extent realized. It should be added, how- rent schooling system and in the educational attainment ever, that discriminatory grants were made to localities, of adults. In 1970 less than 25 percent of adults in the and the ability of local officials to get grants and good 126 urban areas had more than primary school educa- utility service was important. tion (four to five years). The region produces a full range of manufactured products. Its large metropolitan areas are highly diver- Government Influence on Industrial Location sified and produce a wide range of goods and services, while smaller urban areas (under 250,000 population) Under a free market situation in which localities are are highly specialized. Specialization-where, say, 10 fiscally autonomous, certain broad production patterns percent of local employment is in only one industry- for cities can normally be expected to emerge (Hender- occurs in such activities as production of steel, textiles, son 1982). Smaller cities serve agricultural areas, act as apparel, pulp and paper, chemicals, transport equip- administrative centers (state capitals), or engage in ment, and machinery, in food processing, and in ser- manufacturing where processing of raw materials or use vices (export activities in ports and retailing and of heavy processed materials is involved, such as in wholesaling in towns that serve as traditional agricul- production of iron and steel, transport equipment, tex- tural service centers). Almost half of the 126 urban areas tiles, and pulp and paper and in food processing. Manu- are relatively specialized in the production of one manu- facturing centers tend to specialize and to be located factured product. near the materials suppliers, and thus, like the natural Brazil is partially a state capitalist economy. Govern- resources they use, they are spatially dispersed. Smaller ment-owned banks control investment of most domestic urban areas in Brazil do conform to this pattem. Urban Concentration and Decentralization: Brazil 89 Large metropolitan areas may be expected to support balance of payments problems caused by oil price in- clusters of footloose industries-service, high-technol- creases were stemmed by the rapid growth of manufac- ogy, and light manufacturing activities-which find it tured exports. The abandonment of pollution control advantageous to locate in large urban market areas and policies and of the medium-size-cities programs are not tied to the locations of natural resources. But appeared to have occurred in the past partly out of fear Brazilian metropolitan areas, in particular Greater Sao that decentralization of industry out of Greater Sao Paulo, do not conform to this pattern. The reasons for Paulo might lead to failure of the industries and stunt the difference, and its implications for efficiency of re- the growth of manufacturing. The same was true of source usage, need to be explored. longer-term goals of industrial and economic develop- Through its control of utilities, banking, and much ment. The strong belief in the critical role of the eco- steel production, the national government can strongly nomic environment of Greater Sao Paulo persisted in influence and to a considerable extent even determine spite of the success of private manufacturing enterprises the location of heavy manufacturing. It appears that the in cities other than Greater Sao Paulo and Rio. government has so acted and that, within the region That belief essentially rested on the premise that met- under study, it is strongly biased toward locating invest- ropolitan areas offer strong economies of urbanization. ment in Greater Sao Paulo and Rio or on the Greater Sao Economic theory holds that the agglomeration of eco- Paulo-Rio axis. Since the government only recently nomic activity in limited spatial areas occurs because of started to evaluate the overall impact of its past deci- economies of scale in production. At the industry level sions, it is difficult to prove the above contention, but there are urbanization and localization economies. several pieces of informal evidence, some relating to the Urbanization economies have to do with the scale of past, others to the present, indicate a governmental role economic activity in cities in relation to efficiencies in in promoting concentration in the past. urban labor markets, interindustry communication, and First, in a number of well-known cases locational specialized support services and urban infrastructure. decisions were made solely on the basis of governmental Localization economies relate to scale within an indus- decisions-for example, the location of the steel works try as a result of efficiencies in labor markets and ser- in Volta-Redonda and of major power lines and trans- vices specific to that industry and a greater degree of formers. Second, at least in the period under review, the interfirm specialization in detailed activity within the govemment appears to have opposed policies that would industry. lead to decentralization. The pollution control regula- If the economies obtained are localization economies tions that would force some decentralization from there is, from the production point of view, no obvious Greater Sao Paulo have only recently received strong reason for encouraging the location of different unre- government support. The medium-size-cities program lated industries in the same urban area, and so doing is to encourage decentralization has never been effectively inefficient on the consumption side (Henderson 1982). implemented. Third, by offering a high real rate of re- Specialization therefore occurs in smaller urban areas. turn on deposits, the government banks attract a high Of course, if one industry uses inputs from another proportion of national savings (excluding perhaps those industry which are expensive to transport (as in the case of the plantation families). Apart from agricultural activ- for automobiles and steel), they will locate close to- ities, banks appear to be restricted to making long-term gether, although not necessarily in the same urban area. loans only to very large firms (those with more than Economies of urbanization do provide a basis for 1,000 employees). Applications for these loans can only locating different industries together to enhance each be made and pursued in state capitals such as Sao Paulo other's level of efficiency.2 The existence of urbanization and Rio. There is effectively no stock market. Other economies as a result of the creation in Greater Sao long-term investments come primarily from the activi- Paulo of a large-scale economic environment for all ties of international corporations and the large planta- kinds of manufacturing could then be grounds for pro- tion families. moting spatial concentration. To evaluate this issue we If the above appraisal of the spatial bias of government need to test whether the scale economies of the heavy decisions concerning industrial location is correct, manufacturing industries in Brazil pertain to localiza- there remains the question of why the government was tion or urbanization. If there are no economies of urban- biased in favor of Greater Sao Paulo. Political considera- ization, but only economies of localization, it would tions may have been one relevant factor, but there was appear that encouraging or forcing manufacturers to also a widespread belief among the authorities and their pay the relatively high wages and land rents in Greater advisers that Greater Sao Paulo provided an economic Sao Paulo would yield no gain in efficiency. Concentra- environment that was essential to industrial develop- tion would be an error because urbanization economies ment and efficiency. For example, in the 1970, Brazil's are the rationale for encouraging, if not forcing, a con- 90 J. Vernon Henderson centration of different heavy industries in one metropo- The value E is the elasticity of firm output with respect to litan area. Furthermore, other considerations weigh industry employment, holding firm inputs fixed; L is against such concentration, as elaborated in "Conclu- own-industry employment; P is urban population; and , sions and Policy Implications," below. is the elasticity with respect to population. The specifica- tion that e declines as L rises is supported by a quadratic formulation. The interpretation of the e or 1 elasticities Sources and Magnitudes of Economies is that a 10 percent increase in own-industry employ- of Scale ment or in city population, respectively, leads to a 0.1 * e or a 0.1 1 3 percent increase in output for any firm, if the To examine the nature of economies of scale, produc- firm's own inputs are held constant. That is, for the same tion functions and factor ratio equations for different inputs, a firm's output goes up because the external shift industries have been estimated for the region under factor, g(S), is larger. consideration. For each industry the unit of observation Three other problems need to be addressed in the is an urban area. The production function for value estimation of equation 7-3. First, both technology and added is written as labor force may vary among cities. These items are (7-1) X = g(S) X (K, L) controlled by measures of labor force quality, Q, specific (7-1) X g(S) X (K, L ) to an industry in an urban area. Measures of educational where X (0) represents firms' own constant returns to attainment, such as average years of schooling or per- scale (CRS) technology, K and L are inputs of capital and cent of the labor force with three or fewer years of labor, respectively, and g(S) is a Hicks-neutral shift schooling in that industry in that urban area, and mea- factor, external to the firm, whose arguments are scale sures of experience, such as average age, were tested. measures defined at the urban area level. Since X () is Second, X (-) may not be homogeneous of degree one. CRS, on the assumption that all firms in the city face the To test for degrees of homogeneity different from one a same technology, one can aggregate over firms and use measure of average firm size in that industry in that observations on the industry and urban area levels. The urban area was used. The measure was the average external economy formulation and CRs are consistent number of employees per firm, FS. with the assumption of perfect competition (Chipman Third, X should be a measure of physical output, but 1970). The assumptions of Hicks neutrality and CRS to in our formulation the variable is expressed in terms of the firm have been tested, and the results are reported value. The procedure would be satisfactory if output below. prices were invariant among urban areas, but in reality Assuming CRS for X ( , equation 7-1 may be written they are not. We therefore hypothesized that received (7-2) XIL = g(S) X (k) price declines with distance D from the nearest major coastal metropolitan area. Where XI is the measure of where k = KIL. Taking logs, defining log X(k) = f(log value added and X is the true quantity equivalent, it is k), and performing a Taylor series expansion of f(-) about hypothesized that k = 1 yields a translog specification for equation 7-2: X = X*pol (1 - tD) 1 (7-3) log (XIL) = CO + log g(S) + l '(0) In(k) where po is price in coastal metro areas, t is unit trans- + ½f"(0) lln(k)1. *port costs, and D is distance in kilometers. On this basis The arguments in g(S) relate to measures of localiza- it may be approximated that tion economies and of urbanization economies. The (7-5) logXI = logX+ logpo - tD. localization economies are measured by levels of own- industry employment in the urban area; the urbaniza- Using the above result and combining equations 7-3, tion economies are represented either by the urban 7-4, and 7-5, a final estimating equation is obtained: population or by the total employment in the urban (7-6) log(X* IL) = C1 + b1 log(k) + b2 (log k)2 area. Among the various functional forms tried, the best (+) (-_ one, defined informally by such considerations as low multicollinearity and high adjusted R2s, was + b3 FS + b4 Q + b5 D (0) (+) (-) (7-4) g(S) = exp(oxlL) + L logP +b6 (1L)+b7 log P. where (-) (+) d log (X) _ a Expected signs are noted below the coefficients. For the d log L L log(k) terms, the sign restrictions must hold in the Urban Concentration and Decentralization: Brazil 91 neighborhood of k = 1 to have positive and diminishing asks what the firm could sell its equipment, struc- marginal products. tures, and land for today. (Other questions ask The remaining question is whether scale effects are book value and depreciated book value.) Hicks neutral. Combination of marginal productivity FS Average firm size: L divided by number of firms conditions on the basis of equation 7-1 yields the general t Property tax rate: industry property tax payments equation divided by K. This varies by industry and urban (7-7) k = k (w, pL, S) area according to exemptions granted. Q Percentage of labor force with three or fewer years where ia is gross per employee costs (wages plus ben- of schooling, by two-digit industry, calculated efits) in an industry in an urban area and > is the cost of directly from 25-percent long-form sample of capital. It is assumed that either the pretax cost of capital 1970 Demographic Census is the same everywhere or that it increases with distance D Distance in kilometers to nearest coastal port. For D from the nearest major coastal metropolitan area all six ports the urban area is a major metropolitan (which will also be a state capital and the administrative centr fo lon-ter ban loas). he pstta cos ofarea. There Is only one major interior metropol- center for long-term bank loans). The posttax cost of itan area in the sample, Belo Horizonte. Sao Paulo capital varies with the effective local property tax, tp, on is counted as a port although it is 75 kilometers equipment and structures in that industry in that urban from the sea and the actual port is Santos. area. P Ubnpplto If scale effects are Hicks neutral, S should in fact not P Urban population appear in equation 7-7. A measure of own-industry size As a check on the results, a dual form of equation 7-6 was inserted to test whether this is the case. A signifi- was estimated, where cant positive or negative coefficient would indicate non- log (XIL) = F(log w, D, 1IL, P, Q). neutrality. Finally, when firm size and labor force quality are The scale-economy results for the dual form are indis- controlled, the estimated form of equation 7-7 is tinguishable from those for the primary form. (7-8) log k = C2 + a, logw + a2D The sample size for the industries involved was rather +M (_) small, and in the final results presented here some insignificant variables are omitted. The results for equa- + a3tp + a4log L tion 7-6 are given in tables 7-1 and 7-2. (-) (0) + a5 Q + a6FS. Economies of Scale (/) (0) Table 7-1 provides essentially no evidence of urbaniza- Expected signs are noted below the coefficients. tion economies, but there is strong evidence of signifi- Equations 7-6 and 7-8 were estimated by ordinary least squares for three major two-digit industries located Table 7-1. Production Function: Results by Industry in Greater Sao Paulo, namely, iron and steel, transport equipment, and chemicals. The variables are defined as Variable Steel equipzent Chemicals follows: 1IL -109.652 -68.873 -119.224 X Value added (value of production less total mate- (2.01) (1.95) (2.11) rials costs less production taxes). Production tax log(P) -0.019 -0.003 0.091 rates vary spatially, and their differences may not (0.24) (0.04) (1.06) be passed on to consumers. The inclusion of pro- Firm size 0.108 0.166 -0.155 duction taxes in value added has a minimal impact (0.65) (1.78) (1.09) on the results. Education (percent -0.390 -0.310 -0.266 * . I~~~~~~~~~ow-level) (0.41) (0.52) (0.48) L Average monthly number of employees minus (a lowl) 0.443 -0.024 0.160 trivial number) owners and directors. Information (3 51) (0.23) (1.37) on hours worked is not collected. Distance to port 0.082 0.013 0.040 o Total salaries less payments to owners and direc- (1.87) (0.33) (0.92) tors plus firm contributions to social security, Constant 1.677 2.654 2.854 private insurance, and pension programs, all di- Adjusted R2 0.54 0.57 0.50 vided by L Number of cases 36 27 28 K Market value of capital stock. Census question Note. The production function is log (X'IL). 92 J. Vemon Henderson Table 7-2. Own-Industry Scale Effects Table 7-3. The Capital-Labor (KIL) Ratio Transport Fransport Variable Steel equipment Chemicals Variable Steel equipment Chemicals E (median) 0.129 0.153 0.264 log(w) 1.230 0.482 0.703 e (2,000 employees) 0.055 0.034 0.060 (3.09) (1.17) (1.72) e (5,000 employees) 0.022 0.014 0.024 Property tax rate -0.934 - 19.115 -5.098 Note: E, elasticity of firm output with respect to industry employ- (1.54) (4.07) (1.69) ment. log(L) - 0.031 0.013 0.123 (0.28) (0.13) (0.97) cant localization economies. Table 7-2 shows that local- Firm size 0.345 0.079 - 0.016 ization economies tend to peter out by the upper tails of (2.02) (0.49) (0.10) employment (in our samples, 5,000 employees). These Constant - 0.897 1.845 1.389 results indicate that probably little benefit is to be gained Adjusted R 0.43 0.59 0.40 by encouraging the location of these different industries Note: The KJL ratio is In K. in the same urban area to obtain scale economies. There appear to be strong but diminishing benefits from elasticity of substitution in production. The numbers for agglomerating a particular industry's employment in transport and chemicals are in the range of normally the same urban area. accepted values; for steel they are on the high side. Property tax rates have unexpectedly strong impacts Other Factors on use of capital, an indication that the tax significantly distorts investment decisions. The elasticities for the The (log k)2 term is omitted in the final results be- industries, evaluated at average tax rates, are - 0.04 for cause a quadratic effect could not be isolated. Log k is steel, -0.66 for transport, and -0.10 for chemicals. generally significant with the expected sign. The labor Evaluated at the average opportunity cost of capital force quality measures had the correct signs but were (assuming a pretax cost of 0.12), the elasticities are disappointing in their lack of statistical significance, -0.15 for steel, -2.95 for transport, and -0.70 for given their considerable variability among urban areas chemicals. and the fine detail of the data. It could only be concluded Firm size had no impact except in steel. The results that labor force quality is probably not a critical factor in for steel may represent a state capitalism phenomenon, these industries. Firm size performed in no consistent whereby state-owned firms are larger and have access to fashion and was not significant. subsidized capital. Distance to the nearest port had the Distance to the nearest coastal major metropolitan incorrect sign. area was either insignificant or had an incorrect sign. The above results indicate the absence of urbanization This indicates either that output prices are essentially economies and the existence of strong localization econ- spatially invariant or that this effect is offset by an effect omies. For steel, which is 50 percent state owned, there in which efficiency increases with distance from the is a suggestion of general inefficiencies and distortions nearest coastal major metropolitan area. For example, in use of capital. All this points to a gain from relaxation for steel the extent of government control over and of state control over industrial location and production influence on production undoubtedly declines with dis- decisions and thus a gain from permitting decentrali- tance from the nearest state capital and hence efficiency zation. may increase with distance. Scale Biases Environmental Considerations in Decentralization For steel and transport equipment, scale effects are clearly unbiased (table 7-3). For chemicals log L has a Market forces in countries such as the United States, positive coefficient which is weakly significant. Even Canada, and the United Kingdom have tended to pro- that may represent not a scale bias effect but problems in mote the location of the heaviest polluters in smaller aggregating into the two-digit level the capital-intensive cities, away from the largest and most densely populated petrochemical industry, with its high levels of employ- urban centers. In Brazil the worst industrial polluters ment per urban area, and the traditionally spatially dis- remain in the largest population centers. Although air persed chemical sector. Examination of the latter sector quality ought not to be the sole consideration in indus- showed no evidence of scale biases. trial location decisions, Sao Paulo's air quality is abys- Wage levels have their expected strong impact on use mal compared with that of large metropolitan areas of capital. The wage coefficient can be interpreted as the elsewhere; residents of the area view air quality as a Urban Concentration and Decentralization: Brazil 93 Table 7-4. Air Pollution in Greater Sdo Paulo, Compared with U.S. Metropolitan Area (micrograms per cubic meter, unless otherwise indicated) Sulfur dioxide Particulates Days Days Number minimum minimum Metropolitan of Worst Median standard Worst Median standard area stations station station exceededa station station exceededb Los Angeles County 13 53 32 0 164 100 3 New York City 14 115 69 8 82 63 1 Chicago (Cook County) 44 74 28 1 201 85 53 Greater Sao Paulo 13 n.a. 1430 17 n.a. 115c 121 a. Minimum standard is 365 micrograms per cubic meter. b. Minimum standard is 260 micrograms per cubic meter. c. Average. Sources: United States: Environmental Protection Agency, 1977 data; Brazil: Thomas, Comune, and Rizzieri (1980), 1978 data. prime urban problem (see chapter 13). Table 7-4 pre- about decentralization or a different distribution of city sents comparable data for Los Angeles, New York, Chica- sizes for its own sake may not be justified. The provision go, and Greater Sao Paulo, with the emphasis on sulfur of more uniform incentives to middle-size cities-which dioxide and particulates, which are industrial pollutants. could simply mean the elimination of any special incen- The average value for sulfur dioxide for the thirteen tives, direct or indirect, to larger cities in the southern stations in Greater Sao Paulo exceeded by far the value at region-coupled with environmental restrictions in the the worst reporting station among U.S. cities. It would highly damaged and builtup areas, could lead to some seem that encouragement of decentralization of indus- economically beneficial decentralization of activities. A try outside the Greater Sao Paulo valley would have more detailed review of environmental policies is pro- important benefits. vided in chapter 13. The implications of various gov- emmental policies-direct and indirect, intended and unintended-for industrial location and spatial con- Conclusions and Policy Implications figuration of overall economic activities are further analyzed in chapters 10 through 14. This brief account of the Brazilian government's in- fluence over industrial location in the past, against the Notes backdrop of an empirical evaluation of scale economies, brings out some important conclusions. Localization economies are strong, and therefore agglomeration of takes home at least $250 a montv (1980)f plus a full set of firms into specialized cities to take advantage of such takes homefatls $5 a monthi (1980), plus a fUnlet tr . . . . j . . ~~~~~fringe benefits. In comparing this estimate with the United benefits as efficiencies in labor markets and in services States, it should be realized that the cost of nontraded goods is specific to an industry and greater interfirm specializa- much less in Brazil than in the United States. tion within an industry is advantageous. 2. In fact urbanization economies that differ across indus- The results presented here do not show any signifi- tries can also lead to specialization among cities because the cant urbanization economies at the scale of activities optimal size of city for each industry differs. prevalent in the urban centers of the South and South- east in 1970. The rationale for efforts to encourage Bibliography industrialization of the largest urban areas rests on the putative net benefits for heavy industries from locating Chipman, J. S. 1970. "Exteral Economies of Scale and Cor- in areas with a large general scale of economic activity. petitive Equilibrium." Quarterly Journal ofEconomics, vol. The findings of this study do not, however, support this 84, no. 3 (August), pp. 347-85. rationale. Rather, they indicate that efforts to limit or Henderson, J. V. 1982. "The Impact of Government Policies on counter decentralization initiatives are not desirable. Urban Concentration." Journal of Urban Economics, vol. In addition, negative extemalities in the form of en- 12, no. 3 (November), pp. 280-303. vironmental degradation could constitute grounds for Thomas, V., A. E. Comune, and J. Rizzieri. 1980. "Control of promoting some degree of decentralization. As noted Industrial Air Pollution in Sao Paulo: Evaluating Costs, earlier, however, the size distribution of Brazilian cities Benefits and Spatial Effects." University of S&o Paulo. Pro- is by no means excessively skewed, and efforts to bring cessed. 8 Employment Location and Spatial Policies: Colombia and Korea Kyu Sik Lee In the next two decades the urban population in de- of management. The second and perhaps more easily veloping countries is likely to continue to grow about understood policy concern focuses on regional equity, four times as fast as in the industrial countries. Between since in most countries it is politically and socially 1975 and 2000 the urban areas of developing countries important to maintain a balance among regions in in- are expected to absorb about 1 billion people. In the come, opportunities for education and employment, and mid-1970s the net annual addition to the populations of urban amenities. Concerns about regional equity also Mexico City and Sao Paulo, for example, was over a half often underlie programs and policies that attempt to million each; the number was over a quarter million redirect population growth from large cities. each in Jakarta and Seoul. By 2000 the developing world In the middle-income countries of Latin America and will have forty cities with 5 million or more people; East Asia, including Mexico, Brazil, Venezuela, the Phil- twenty are expected to have more than 10 million peo- ippines, and Korea, policymakers have paid a good deal ple. Until 1950 Buenos Aires was the only city in the of attention to various ways of reducing spatial biases developing world with a population of over 5 million toward large cities. More recently, the govemments of (World Bank 1979, p. 72). those countries have initiated specific policies to decen- The rapid urbanization in these countries has pro- tralize economic activity away from the largest urban duced a heavy concentration of population and eco- center to peripheral areas or secondary cities. Policy nomic activity in a very few large urban centers. This packages using diverse instruments have been initiated pattern of urban concentration has generated two main and implemented with varying degrees of success (Re- policy concerns. First, it is widely believed that the naud 1981; Townroe 1979). Industrial location policies largest cities in these countries are getting "too big." tend to be the most important of the decentralization This belief is not usually based on evidence that negative policies. This is not surprising, since "the true determi- extemalities such as pollution and congestion are nants of urbanization and spatial concentration in de- greater than the benefits of agglomeration economies. It veloping countries are found in the forces that deter- is more likely that the concems about the size of the mine the location of employment opportunities: the large cities stem from the decline in the quality of life of nature and pattern of industrialization, the pace of agri- their high-income groups, from the frustrations of plan- cultural development, and the growth of transportation ners who have experienced enormous pressure in recent and communication networks" (World Bank 1979, years to accommodate rapid urban growth, and from a p. 76). More specifically, the decentralization policies fear that large cities may experience catastrophic failure include explicit policy instruments that are intended to relocate existing industries from the large urban centers or to induce new industries to new industrial zones, Note: This chapter has been published in somewhat different form secondary cities, or lagging regions. as "Decentralization Trends of Employment Location and Spatial Policies in LDc Cities," Urban Studies, vol. 22 (April 1985), pp. 15142. The rationale for policies to decentralize economic It has been edited slightly. activity is difficult to justify solely on economic grounds 94 Employment Location and Spatial Policies: Colombia and Korea 95 because these policies are an outcome of diverse social, tion on firms where the respondents worked. The ques- political, and economic objectives: interregional equity, tions covered the location of the workplace, the number political cohesion, national defense, and prevention of a of workers in the firm, the initial year of operation at the catastrophic failure of urban management. Because de- present location, the previous location (if applicable), centralization policies are deemed plausible responses to and the type of business. this complex set of policy objectives, they are likely to be Making inferences about the characteristics of firms pursued in the middle-income countries, and it is ex- where the household members work on the basis of a tremely important to help governments select policies household survey requires some justification. The that are least damaging to the overall welfare of the household survey results can be compared ex post with economy. the social security data regarding firm characteristics For analytical reasons it is convenient to address em- important for this study, because the household survey ployment location policies at two levels of spatial questionnaire asked the respondent whether his firm aggregation: deconcentration policies that are intended was affiliated with the social security system. The dis- to modify employment location patterns within the tributions given by the household survey regarding the capital region and decentralization policies that are in- firms affiliated with the social security system can there- tended to influence spatial patterns at the national level. fore be compared with the corresponding distributions The former involves a study of intrametropolitan phe- from the social security data. In terms of the percentage nomena; the latter deals with interregional issues, in- distribution of several firm characteristics-size, type of cluding the development of secondary cities and lagging industry, and location-the two data sets are strikingly regions. This chapter is confined to intrametropolitan similar, which increases our confidence in inferences issues. drawn from the household survey. Most of the analyses To formulate sound spatial policies and to implement in this section are thus based on the information from plans and programs efficiently, it is essential for policy- the 1978 household survey, which covers all workers makers to understand the trend of employment location regardless of the size of the firms in which they work. patterns and how firms' location behavior tends to pro- Employment shares of major sectors in Bogota and duce a trend. In the Bogota City Study conducted by the Cali are similar to those of other cities of the world. As World Bank, analytical and empirical results were estab- shown in table 8-1, the share of manufacturing employ- lished regarding firms' location decisions, but a study of policy effects was not undertaken because explicit policy instruments were not implemented in BogotA. Measure- Table 8-1. Employment Share of Major Sectors ment of such policy effects is the focus of World Bank in Selected Cities research on employment location policies in Korea, a (percent) country with a long history of policy experiments. This Manu- chapter reports the empirical findings from the Bogota City facturing Commerce Servicesa City Study, which had a heavy behavioral focus, and Bogotg omeceebie draws some policy conclusions from the Korean experi- (1978) 23.6 20.3 40.9 ence with spatial policy. Cali, Colombia (1978) 31.1 21.6 32.6 Observed Patterns of Employment Location Seoul, Rep. of Korea (1970) 22.7 28.9 33.0 This section documents the employment location pat- Kuala Lumpur, terns of four major industry groups-manufacturing, Malaysia (1970) 20.5 17.6 35.5 commerce, finance, and services-which together Manila, Philippines account for about 85 percent of total employment in (1970) 22.1 13.8 37.3 Bogota and Cali.' Three main data sets are used: the 1978 Abidjan, household survey for both cities, the 1972 household C6te d'Ivoire (1970) 22.5 18.6 42.8 survey for BogotA, and the social security establishment Tu1i9s 18.9 17.5 39.2 files for Cali (1976) and Bogota (1978). Ideally the data United States, for the study of employment location should be based on small cities (1960) 25.1 28.4 28.8 the establishment. Although the social security files are United States, such data sets, their coverage of small firms is poor. The large cities (1960) 30.2 32.6 24.8 1978 household survey, which was jointly prepared by a. Includes financial services. the World Bank and the Colombian National Statistics Sources: The values for Bogota and Cali are based on the 1978 DANE Department (DANE), had a module for obtaining informa- household survey; those for other cities are from Renaud (1981). 96 Kyu Sik Lee ment is remarkably consistent among developing- Map 8-2. Cali: Ring Systen Based on Comunas country cities (from 20 to 25 percent). Bogota's share of manufacturing employment falls in this range, but Cali's share, 31 percent, is similar to that of large U.S. cities. The employment shares in commerce-about 20 per- cent for both Bogota and Cali-are comparable with those of other developing-country cities. Bogota's share Comuna boundaries of employment in services (41 percent, including Ring boundaries finance) is on the high side and Cali's 33 percent is on the low side, compared with other cities in developing countries. Bogota's position as the nation's capital and 3 financial center may contribute to its high share of services employment. To study the spatial distribution of employment, we have devised a ring system for each city based on comu- nas, the administrative units in Colombian cities. Bogo- ta has thirty-eight comunas and Cali has twenty-eight. By aggregating comunas we divided Bogota into six rings and Cali into five (maps 8-1 and 8-2). The ring 4 system is useftil for studying the relation between the spatial distribution of employment and distance from 4 the central business district (CBD). Table 8-2 shows the spatial distribution of employ- \ ~ ~~5 Map 8-1. Bogota: Ring System Based on Comunas Comuna boundaries Ring boundaries ment by ring for major industry groups in Bogota and Cali. For all employment, Cali shows a greater central < ? g 6 1 5 tendency than Bogota; the peak concentration is in ring 3 in Cali and in ring 5 in Bogota. In Cali the share of employment drops markedly beyond ring 3. The CBD'S share of employment is not much different in the two cities-14 percent in Bogota and 16 percent in Cali, which is comparable to that of large U.S. cities.2 The central tendency is, however, substantially different ¢< y 4 4 ~; among industry groups. In both cities the concentration ) \ X } 13 S of employment in the central area is highest in finance, followed by commerce, services, and manufacturing. For all industry groups, however, the central area in Cali 2 has a much larger share of employment than that of <, \/ / 1 ev Bogota. To examine the changing patterns of employment location, the spatial distribution of employment for at least two points in time must be examined. For the trend analysis for Bogota two household survey data sets are used: the 1972 household survey conducted for the Bogota Urban Development Study, Phase 11, and the 1978 DANE household survey. Like the 1978 survey, the Phase II survey contained questions about the respon- dent's workplace, including location, firm size, and type Employment Location and Spatial Policies: Colombia and Korea 97 of business. For Cali the social security data file for 1976 commerce. The service sector also shows a clear trend of and the 1978 DANE household survey were used. employment decentralization. The location pattern of Table 8-3 reports the spatial distribution of employ- financial establishments is different from those of other ment in Bogota by major industry group for 1972 and sectors in that the central tendency persisted during the 1978. We find evidence of decentralization of total em- six-year period. For this sector, however, it should be ployment during the period, but the sharp decline of the noted that ring 2 had a substantial gain in employment CBD'S share (from 23 to 14 percent) is suspect. The share, which reflected the shift of employment to the apparent overstatement was largely attributable to the International Center (see note 3). problem with the definition of the CBD in the 1972 Table 8-4 shows evidence of employment decentra- survey. It should be noted that the share of the CBD and lization in Cali on the basis of the social security data for ring 2 together decreased only slightly, from 36.6 per- 1976 and the 1978 household survey data. As indicated cent in 1972 to 31.7 percent in 1978.3 earlier, there is a strong consistency between the social In BogotA the extent of employment decentralization security data and the subset of the 1978 household in manufacturing is very similar to that in commerce, survey data which represents those individuals whose except that the CBD kept a larger share of employment in firms are affiliated with the social security system. Since Table 8-2. Employment Distribution, by Ring and Major Industry Group, 1978 (percent) All Ring employmenta Manufacturing Commerce Finance Services Bogotd 1 (CBD) 13.95 6.01 15.75 41.43 12.91 2 17.74 13.47 19.77 29.38 18.68 3 16.40 21.54 14.83 11.43 16.87 4 20.60 24.89 19.37 10.88 23.18 5 24.94 28.25 27.72 5.65 21.79 6 3.43 2.19 1.51 0.60 4.41 Location not specified 2.96 3.66 1.05 0.63 2.15 Cali 1 (CBD) 16.36 9.83 25.23 67.41 14.09 2 26.16 25.71 23.68 19.69 34.57 3 32.67 39.64 31.64 4.70 30.24 4 14.37 11.81 11.84 1.57 15.26 5 2.78 1.38 4.53 0.00 2.12 Location not specified 7.66 11.62 3.08 6.64 3.73 a. Includes other sectors. Source: DANE household survey, 1978. Table 8-3. Changes in Employment Location, Bogotd, 1972-78 (percent) All employment' Manufacturing Commerce Finance Services Ring 1972 1978 1972 1978 1972 1978 1972 1978 1972 1978 1 (cBD) 23.03 13.95 18.20 6.01 19.43 15.75 42.11 41.43 22.62 12.91 2 13.61 17.74 16.07 13.47 12.18 19.77 13.69 29.38 12.74 18.68 3 14.62 16.40 18.94 21.54 13.35 14.83 6.89 11.43 15.88 16.87 4 18.80 20.60 20.27 24.89 21.83 19.37 10.00 10.88 20.74 23.18 5 18.61 24.94 21.76 28.25 21.52 27.72 14.64 5.65 17.89 21.79 6 1.67 3.43 1.04 2.19 3.42 1.51 0.55 0.60 1.88 4.41 Location not specified 9.67 2.96 3.72 3.66 8.27 1.05 12.12 0.63 8.25 2.15 a. Includes other sectors. Sources: DANE household survey, 1978; Bogota Urban Development Survey, Phase 11, 1972. 98 Kyu Sik Lee Table 8-4. Changes in Employment Location, Cali, 1976-78 (percent) All employmenta Manufacturing Commerce Finance Services Ring 1976 1978 1976 1978 1976 1978 1976 1978 1976b 1978 I (CBD) 31.51 26.19 20.19 14.94 48.38 54.97 45.68 84.90 29.15 25.34 2 37.12 27.61 34.28 27.02 34.50 24.40 16.67 15.10 55.47 32.19 3 28.40 31.33 41.40 46.27 11.09 15.65 37.55 0 14.79 16.45 4 2.95 13.82 4.09 11.03 6.03 4.97 0.11 0 0.54 24.37 5 0.03 1.05 0.05 00.70 0 0 0 0 0.06 1.65 Note: Based on employment in firms with ten or more employees. a. Includes other sectors. b. Excludes employment in the government sector. Sources: DANE household survey, 1978; social security data files, 1976. the social security files' coverage of small firms for all The results reported in table 8-5 indicate that in both industry groups is poor, to increase the comparability Bogota and Cali the CBD experienced a net loss of jobs in between the two data sets table 8-4 includes only jobs in each industry group. The net outflow of jobs from the firms with ten or more employees. CBD was much greater in Bogota than in Cali for each For all industry groups together there is clear evi- industry group except the service sector. In Cali all rings dence of employment decentralization in Cali, as in outside the CBD had net gains of jobs except for service Bogota. The results reported in table 8-4 indicate that employment in ring 3. Although Cali's ratio values sug- although Cali experienced employment decentralization gest an outflow from the CBD to the other rings, within in manufacturing and services, there is an increasing non-cBD rings the values do not indicate decentraliza- central tendency in commerce and finance. The sharp tion. increase in the csD's share of employment in finance For Bogota, the results show evidence of decentraliza- during this period is obviously a result of sampling tion within the non-cBD rings, owing to relocation pat- errors: the 1978 household survey must have under- tems. In Bogota's manufacturing the ratio gradually sampled individuals working in financial firms, espe- declines as the distance from the cBD increases; rings 4 cially those located in the outer rings. Nevertheless, it is and 5, for instance, had a net gain in employment. The reasonable to conclude that these results do not support gain of jobs in ring 2 is also reflected in table 8-5 and a decentralization trend in these two sectors in Cali. indicates the growth of the International Center. The trend of employment decentralization having Table 8-6 reports the location pattems of jobs created been examined, we now analyze the factors that in- by firms established during the five-year period (births). fluence changing patterns of employment location by The strong regularity in the location patterns of new investigating the location behavior of four groups of jobs is striking; in both Bogota and Cali the employment firms: firms that relocate (movers), newly established share of new jobs by ring increases with the distance firms (births), defunct firms (deaths), and firms that from the CBD for almost all industry groups except remain at the same location but expand or contract. finance. This is an indication that newly established Evidence will be presented for the first two categories. firms tend to locate in outer areas and thus contribute to The destinations of movers and the location patterns of the decentralization patterns. It is worth noting that in births are important for understanding the changing Bogota's finance sector 43 percent of jobs created by new patterns of employment location and for predicting the firms during 1973-78 were located in ring 2, a result future spatial structure. Here we use the relevant in- that captures the northward shift of financial activities fonnation collected in the 1978 household survey to from the CBD to the International Center. examine the direction of job movements and the loca- tion patterns for births. The 1978 household survey asked when the firm be- Behavioral Underpinnings gan operations at its present location and what was the of Location Choice in Manufacturing previous location if it had moved. On the basis of this information the origin and destination ratio was calcu- The previous section summarized, with the use of lated for each industry group with the rings as subareas. various data sets, the location patterns of employment in The ratio was obtained by dividing the number of jobs Bogota and Cali and their changes over time. The results moving out of an area by the number of jobs moving into reveal a high degree of employment location dynamics that area during 1973-78.4 and strong evidence of the spatial decentralization of Employment Location and Spatial Policies: Colombia and Korea 99 employment in both cities. To understand and explain changing location patterns of manufacturing employ- these changing location patterns, in 1978 we conducted ment, were reported in Lee (1982a),' and a model of a survey of manufacturing establishments in Bogots employment location was formulated and estimated us- using the DANE industrial directory as the sample base. ing this survey data (Lee 1982b). The remainder of this The survey questionnaire was designed to take no more section is based on Lee (1982b). than one hour to complete and did not require the In this study the basic premise for modeling employ- respondents to refer to their accounting books, but it ment location is drawn from well-established theoretical nevertheless yielded a large amount of information with and empirical foundations in the literature on housing nearly 300 computer-readable variables. The survey re- and residential location. Mills (1972), Solow (1972), and sults. which provide the basis for understanding the Muth (1969) provided the basic theoretical foundations Table 8-5. Moving Patterns of Jobs, Measured by Origin-Destination Ratio for Rings, 1973-78 All Ring employmenta Manufacturing Commerce Finance Services Bogotd I (CBD) 2.37 2.51 3.15 3.32 1.45 2 0.66 1.54 0.45 0.33 0.87 3 1.33 1.58 1.32 1.04 1.36 4 0.55 0.76 0.51 0.06 0.58 5 0.44 0.32 0.20 2.70 0.35 6 0.59 _b _b, c _b, c 0.29 Cali 1 (CBD) 1.85 1.97 1.71 1.58 2.03 2 0.68 0.70 0.61 _c 0.45 3 0.86 0.93 0.84 _b,c 1.21 4 0.67 0.70 __b __b c 0.33 5 3.44d _b _c _b, c _b Note: The origin-destination ratio is found by dividing the number of jobs moving out of a zone by the number of jobs moving into the zone. a. Includes other sectors. b. No firms moved into the area. c. No firms moved out of the area. d. Even though the ratio was high, ring 5 lost only 651 jobs and gained 189, whereas the CBD lost 9,479 and gained 5,126. The total number ofjobs relocated was 28,170 in Cali and 84,755 in Bogota. Source: DAE household survey, 1978. Table 8-6. Location Pattems of New Jobs, 1973-78 (percent) All Ring employmenta Manufacturing Commerce Finance Services Bogotd 1 13.78 7.55 14.09 29.20 18.61 2 17.80 13.47 20.31 43.04 14.05 3 14.42 19.98 8.15 13.86 17.38 4 20.87 24.66 23.00 9.43 17.55 5 27.67 30.62 31.97 4.47 26.78 6 2.64 2.82 1.72 0.00 2.32 Location not specified 2.82 0.90 0.75 0.00 3.32 Cali 1 14.70 9.27 13.61 0.00 26.27 2 25.62 19.18 24.05 60.00 30.88 3 38.33 44.95 44.76 20.00 32.99 4 13.40 22.24 11.45 0.00 5.16 5 2.60 2.29 1.97 0.00 0.00 Location not specified 5.35 2.06 4.15 20.00 4.70 s. Ineludes jik*ar sectlolrs. .8.,urre: DANk. holuseholld survey, 1978. 100 Kyu Sik Lee ! for the economic analysis of residential location and public goods include the quality of public utility services employment location in urban areas. Theoretical and such as electricity and water, the quality of municipal empirical work on housing and residential location has services such as police and fire protection, and zonal advanced rapidly,5 but comparable work in the field of characteristics such as air quality and congestion. employment location has not followed. Mieszkowski and The relevant cost components in calculating the opti- Straszheim (1979, p. xiii) observe that "among the most mum combination of inputs and hence in selecting a important issues not covered in this volume is the rela- particular location are, then, wages, capital cost, input tive importance of different factors in employment loca- materials costs, delivery costs of inputs and output, and tion and decentralization patterns within metropolitan land rent. Following the standard theory in urban eco- areas, which deserves far more research. Principally nomics, a particular plant site is occupied by the firm because of data problems, little research has been con- which bids the highest price for it. The bid price depends ducted on employment locations or urban labor markets on the attractiveness of the particular site (which in turn in recent years. Effective use of the censuses of employ- depends in part on the availability of local public goods) ment is plagued by disclosure rules." Nevertheless, work for a firm of the particular type. In locational equilib- by Leone (1971), Schmenner (1973), and Kemper rium all firms in an urban area make the same profits (1973), followed by Struyk and James (1975), pushed and no firm has any incentive to relocate. This equilib- back the frontier in developing an empirical basis for rium configuration is an outcome of cost tradeoff cal- employment location study. Moreover, recent efforts in culations by individual firms, given the spatial variations actual modeling of employment location and econo- of relevant costs. For example, a large manufacturing metric estimation of such models have been promising: firm may choose a site in a low-rent area near the Hanushek and Song (1978) developed a framework for periphery to meet its need for more plant space, analyzing the spatial structure of employment in the although delivery distance will be greater. Small firms Boston metropolitan area, Erickson and Wasylenko may prefer a central location where the high rent is (1980) estimated a model of relocating firms in the more than offset by externalities available for production Milwaukee metropolitan area, and Schmenner (1973, and shipment. The high rent in the central area can also 1982) provided evidence from his econometric work on be explained by the high intensity of variable input use. Cincinnati and New England. Carlton's work (1977, The empirically observed rent gradient reflects the in- 1979) focused on the intercity location decisions of new creasing marginal productivity of land as the distance to firms. the CBD becomes shorter. Residential location studies usually assume that a The stochastic specification of the model should pro- consumer chooses a particular dwelling unit at a par- vide a framework for predicting the probability that a ticular location at which he maximizes his utility, given firm of a particular type will occupy a site with particular the budget constraint. In a similar optimization attributes (Z). Since a given site will be occupied by the framework, it is assumed that the firm, as a price taker, firm with the highest bid, the relevant random variable locates where it maximizes profits. In making the loca- for determining the probability is the maximum bid tion decision, the attributes of the particular plant site made within a group of firms with similar attributes. and the lot size enter into the process of determining the The probability distribution of a random variable associ- optimum combination of inputs for production. This ated with the maximum bid leads to a multinomial logit means that the site attributes enter directly into the specification for the firm location model. firm's production function in the following way: The application of the multinomial logit method to (8-1) Q = f(L, X; Z) economic research became popular with McFadden's work on travel demand studies (1973, 1976). Such a where Q is output, L is lot size, X is a vector of variable model was used to predict an individual's choice of travel inputs such as labor and plant and equipment, andZ is a mode among a finite number of alternatives-car, bus, vector of exogenous site characteristics. taxi, transit, or foot-given the characteristics of the The site characteristics are independent of lot size and individual or household. Subsequently, the multinomial represent local public goods available to that particular logit framework was applied to housing and residential location. In a recent work on housing Burstein (1980) location studies by, for example, Friedman (1975), Ler- explicitly introduces local public goods into the consum- man (1977), and Quigley (1976). These studies applied er's utility function. Ellickson (1981) integrates the the travel mode choice model to variations of residential theory of residential location choice with that of local choice models: a utility-maximizing consumer of type t public goods by including in the consumer's utility func- chooses a house of type Z, which is analogous to the tion the attributes of the surrounding neighborhood and consumer's choosing a travel mode Z among several the public schools. For studies of firm location, local alternatives. Employment Location and Spatial Policies: Colombia and Korea 101 In his recent work Ellickson (1977, 1981) makes an requires the selection of particular types of industries to important departure from the above approach and offers occupy sites that have particular attributes. Hence, it is an alternative multinomial logit specification for the important for policymakers to understand the require- residential choice model using the bid-rent theory. ments of firms for attaining equilibrium at new loca- According to Ellickson (1981, p. 63), "the most natural tions and to be able to assess the level and costs of way to interpret such models is in terms of a prediction government subsidies and infrastructure investment of what sort of consumer is most likely to occupy a house needed to meet such requirements. with a specified set of characteristics"-the probability The Bogota study described earlier did not test the that a house with characteristics Z will be occupied by a effectiveness of explicit policy instruments, partly be- household of type t.6 Ellickson points out that his cause such instruments were not implemented in that method has several advantages, such as specifying en- city. The behavioral underpinnings established in the dogenously the properties of the disturbance terms (as study, however, provide sufficient clues regarding those the Weibull distribution). policy instruments that are most appropriate for in- Ellickson's approach is particularly relevant for mod- fluencing the location choice of particular types of firms eling employment location. As discussed above, our and for influencing aggregate location patterns. problem is to predict the probability that, given a site The estimation results reported in Lee (1982b) show with particular characteristics (Z), a firm of particular which independent variables (site attributes) are most type will occupy that site. Such a site may exist at any important for the location choices of firms of particular location in an urban area, and combinations of site types. In table 8-7 the independent variables are ranked characteristics can produce a large number of site varia- by order of importance for each type of firm according to tions. The application of the McFadden approach re- the values of the elasticity coefficients. For small firms quires a definition of a finite number of locations (or in both the textile and the fabricated metal industries, subareas) from which a firm of a particular type is accessibility to local input markets (INPUTBT), the com- supposed to choose. Defining alternative subareas in muting distance for production workers (wKsouTH), and that fashion is arbitrary, and the subsequent analysis the population density of the zone of plant location will tend to be descriptive rather than analytical and to (POPDENS) are important site characteristics. These three suffer from the idiosyncracy introduced by the arbitrary variables alone carry more than half of the total coef- definition of alternative locations. ficient weight in the equation. Accessibility to local product markets (PRODSOLD) was the second most impor- tant variable for small fabricated metal firms, but it was A Framework for Evaluating not as important for small textile firms. Policy Effects For large textile establishments the location quo- tient (LOCQT, a measure of scale economies in the zone of A study of the effects of employment location policies requires a theoretical and empirical framework for test- ing the effectiveness of policy instruments with respect Table 8-7. Independent Variables in Firm to the firm's location choice. The model described above Location Choice, by Rank, Bogota provides such a framework. It predicts the probability Small that a firm of a particular type will occupy a site with Small fabricated Large particular attributes, and the site attributes are the inde- Variable textile metal textile pendent variables that are subject to policy manipula- INPUTBT 1 1 4 tions. The attributes include measures of access to out- WKSOUTH 2 4 7 put and input markets, employees' commuting distance, POPDENS 3 3 8 public services, zonal amenities, and external (or scale) ELECINT 4 8 2 economies. This means that the model can predict the LOCQT 6 2 5 incremental effects of alternative policy instruments on ADMNORTH 7 5 3 the likelihood that a firm will choose a particular site. DISTCBD 8 6 6 Policymakers in developing countries often attempt to relocate industries from the traditional industrial Note: INPUTBT, a measure of accessibility to local input markets; WKSOUTH, commuting distance for production workers; POPDENS, districts of a large city to outer areas or to smaller cities. population density of the comuna where the establishment is located; The government's plans may include developing new ELECINT, frequency of electricity interruption; LOCQT, location quotient; industrial towns or estates or expanding existing ones to PRODSOLD, a measure of accessibility to local product markets; ADMNORTH, commuting distance for administrative workers; DISTCBD, induce new or relocating firms to settle in a desired area. airline distance from the CBD. In all cases implementing such plans and programs Source: Lee (1982b). 102 Kyu Sik Lee plant location), the quality of electricity supply occurs because of increased population and incomes, (ELECINT), and the commuting distance for administra- often result in economic inefficiency, since transport tive workers (ADMNORTH) are the three most important improvements have reduced the central city's compara- site attributes; together they carry more than two-thirds tive advantage for production and other economic activi- of the total weight in the equation. Access to local ties. markets and population density are relatively unimpor- Large U.S. cities also have municipal fiscal problems tant for this group of firms. that encourage decentralization. The high per capita tax The results for large textile firms are consistent with burden in central cities in relation to the suburbs pro- the findings that large firms are export-oriented and vides incentives to move to the suburbs. It has been tend to locate near the periphery, where land needed for observed that "the lower the average income level of the modem facilities is available at lower cost and highway central city relative to its suburbs, the smaller is the access is better than in the central area (Lee 1982a). The central city's population, and the larger is the land area evidence on small firms' local market orientation and occupied by the urbanized area" (Muth 1969). Evidence their tendency to locate in high density areas is consis- from Bogota and Cali, however, does not suggest an tent with the incubator hypothesis, which was supported increasing concentration of low-income population in by a test in an earlier work (Lee 1981). The hypothesis the central city, and there is no sign of central city decay. states that small new manufacturing firms tend to locate In fact, policies to decentralize economic activity from in centralized areas that provide essential services for large cities in developing countries stem mainly from production and sales activities. The specific areas in the increasing concentration of economic activity in the Bogota that were identified as incubator areas are not in central city and the perceived problems of congestion the industrial district but are adjacent to the CBD. and pollution as the city's population grows rapidly. From this analysis it is apparent that government Also, municipal fiscal relationships in developing coun- policies intended to influence employment location pat- tries are different from those in the United States. In tems can be effective if such policies influence the site Korea, for example, electric utility service charges, attributes which are important to firms. The analysis which are uniform nationwide, tend to subsidize central further suggests that different mixes of policy instru- cities over outlying areas and hence encourage central- ments should be considered for different groups of firms ization. and industries. The case of Korea affords a chance to look at a con- scious decentralization policy. During the past decade various spatial policies to control the growth of Seoul Policy Implications and to disperse its population have been implemented. In 1971 the greenbelt surrounding Seoul was estab- The strong decentralization of employment in Co- lished. The 1977 Industrial Location Act in effect pre- lombian cities, summarized in "Observed Patterns of vented new manufacturing firms from locating within Employment Location," above, is comparable to the Seoul and enabled the govemment to issue relocation trend observed for large U.S. cities during the past sev- orders to establishments already set up there. That same eral decades (Hoover and Vemon 1959; Leone 1971). year the government initiated a ten-year comprehensive But the main goal of spatial policy in the United States plan for redistribution of population and industry from regarding intrametropolitan decentralization has been Seoul. The plan included many policy instruments in quite different from that in developing countries. In the five principal categories: the relocation of industries United States the main policy objective has been to from Seoul; inducements for people to relocate to south- reduce urban decentralization in the hope of preventing em provinces; the decentralization of educational facili- central city decay. This policy objective is based on the ties; the relocation of various urban functions within the belief that decentralization occurs because of deteriorat- capital region and the improvement of city plan imple- ing conditions in the central city; there is, however, mentation, particularly in dealing with clandestine little empirical evidence to support this view. Muth housing construction in Seoul; and tax and credit incen- (1969) attributes most decentralization to transport im- tive schemes to aid relocating firms. Most of the instru- provements, suburban (new town) development pro- ments stipulated in the plan address the location and grams, and federal subsidies for homeownership. Muth's relocation of manufacturing establishments. A compre- findings suggest that federal programs intended to re- hensive review of employment location policies in Korea duce decentralization, such as urban renewal programs appears in Choe and Song (1982), and a theoretical that tend to lower housing density in the central city, analysis of policy efficiency is discussed by Murray have expedited the trend rather than reduced it. (1982). The empirical evaluation of such policies was the Attempts to reverse the tide of decentralization, which main task of a recently completed World Bank project.! Employment Location and Spatial Policies: Colombia and Korea 103 Several other recent policy measures in Korea and Co- policies on economic grounds has not been established, lombia are discussed below. and little is known of their effects or their welfare im- The incubator hypothesis mentioned earlier states plications. The key policy question is how to guard that small new manufacturing firms start in central against spatial policies that are excessive in relation to locations that provide needed services and infrastruc- prevalent trends, since excessive measures may result in ture and then move to less central locations as they grow serious welfare losses. The lack of empirical information and require more space for expansion (Hoover and Ver- on decentralization8 and policy effects in developing non 1959). This hypothesis was tested for Bogota on the countries does not yet permit the formulation of more basis of industrial directory data (Lee 1981). Data from efficient spatial policies, but policies to decentralize pop- the establishment survey conducted for the Bogota City ulation and economic activity are probably not good Study do in fact indicate that small firms start up in substitutes for better internal management of city central locations (Lee 1982a), and the logit results growth. For example, the effect on air pollution or on summarized earlier also support this hypothesis (Lee traffic congestion of reducing the population or employ- 1982b). There is evidence that Korean firms also follow ment in a large city by a certain amount is likely to be this pattern (Meyer 1981, Lee 1985). The 1977 Industrial small (Henderson 1980; Tolley, Graves, and Gardner Location Act, which prohibits new manufacturing firms 1979). within the city limits, should have restricted such in- cubation. A related issue is whether incubation can be replicated in new industrial estates or existing towns Notes outside Seoul. In 1978, as an important measure for decentralizing - 1. The material in this section is from Lee (forthcoming). population and economic activity away from Seoul, the 2. According to Bronitsky and others (1975), about 10 to 15 government established a new industrial town, Banweol, percent of total employment in large U.S. cities is in the CBD. less than 30 kilometers distant. Although nearly 1,000 3. Moreover, on the basis of the two data sets, the total plant sites for small and medium-size firms were pre- tional Center (comuna 81, the ahrea g dirith thore in the Interna pared, the occupancy rate was less than 20 percent as of stayed almost constant during the six-year period. 1981. Many firms that moved to Banweol suffered excess capacity and financial losses as a result of overinvest- Comuna 1972 1978 ment in plant and land area (induced by the government 31 (CnD) 201,975 166,878 incentive schemes), increased operating costs after re- location, and the general recession of 1980. Total 226,762 229,443 A case study on Banweol (Choe and Song 1982) shows 4. Since the number of workers at the previous locations is that the most serious problems facing the relocated estimated on the basis of the number of survey respondents, it firms have been reduced accessibility to product markets is implicitly assumed that the number of jobs at the previous and input suppliers, the unavailability of production locations is the same as that at the 1978 locations. This workers, and difficulties in obtaining day-to-day busi- assumption should not affect the conclusions of this study. ness information because of poor telephone service and 5. Majorwork in this area includes Kain and Quigley (1975), nimites infsormationbecauseof pontacts.Poortelep e Sere a Straszheim (1974), and Ingram (1977). limited person-to-person contacts. Poor access to Seoul 6. The mathematical derivation of this result can be found and Incheon is largely responsible for these problems. A in Lee (1982b). related problem is the reluctance of production workers 7. "An Evaluation of Industrial Location Policies for Urban to relocate to Banweol or to commute from Seoul. Attri- Deconcentration," RPO 672-58, RPO 672-91. See World Bank, tion of skilled workers has been high, and it is difficult 1984, Abstracts of Current Studies 1984, Washington, D.C. for firms to replace those who quit. 8. Lee (1985) shows that in the 1970s Seoul's CBD experi- That such a seemingly short distance thwarted the enced an annual net loss in manufacturing employment of 7.6 development of Banweol is striking. The logit analysis percent, while the periphery of Gyeonggi province, in which used to study the Bogota data helps explain the Korean Seoul is located, gained employment at a rate of 34 percent. experience: small and medium-size firms prefer central locations, and accessibility to local markets and proxim- ity to production workers are the most important site Bibliography attributes for them. In contrast to the bias against decentralization in the Bronitsky, L., andothers. 1975. Urban DataBook. U.S. Depart- United States, several developing countries have tried to ment of Transportation, Report DOT-TSC-OS5-7545-1. decentralize economic activity away from the central Washington, D.C.: Government Printing Office. city. Nevertheless, the desirability of decentralization Burstein, Nancy R. 1980. "Voluntary Income Clustering and 104 Kyu Sik Lee the Demand for Housing and Local Public Goods." Journal World Bank, Washington, D.C. Forthcoming in Urban of Urban Economics, vol. 7, no. 2 (March), pp. 175-85. Studies. Carlton, D. 1977. "Models of New Business Location." Uni- - . Forthcoming. "Employment Location in Colombian versity of Chicago, Center for Mathematical Studies in Busi- Cities." Water Supply and Urban Development Department, ness and Economics, Report 7756. Chicago. World Bank, Washington, D.C. Processed. -. 1979. "Why New Firms Locate Where They Do: An Leone, R. 1971. "Location of Manufacturing Activity in the Econometric Model." In William Wheaton, ed., Interre- New York Metropolitan Area." Ph.D. dissertation. Yale Uni- gionalMovements andRegional Growth. Washington, D.C.: versity, New Haven, Conn. Processed. Urban Institute. Lerman, S. R. 1977. "Location, Housing, Automobile Own- Choe, S. C., and B. N. Song. 1982. "An Evaluation of Industrial ership and Mode to Work: A Joint Choice Model." Trans- Location Policies for Urban Deconcentration in the Seoul portation Research Board Record, no. 610, pp. 6-11. Region." Urban Development Discussion Paper 7. World Washington, D.C.: Transportation Research Board. Bank, Washington, D.C. McFadden, D. 1973. "Conditional Logit Analysis of Qualitative Ellickson, B. 1977. "Economic Analysis of Urban Housing Choice Behavior." In P. Zarembka, ed., Frontiers in Econ- Markets: A New Approach." R-2024-NSF. Santa Monica, ometrics. New York: Academic Press. Calif.: Rand Corporation. . 1976. "The Theory and Practice of Disaggregate De- . 1981. "An Alternative Test of the Hedonic Theory of mand Forecasting for Various Modes of Urban Transporta- Housing Markets." Journal of Urban Economics, vol. 9, no. tion." Prepared for the Seminar on Emerging Transporta- 1 (January), pp. 56-79. tion Planning Methods. U.S. Department of Transportation, Erickson, R.A., and M.Wasylenko. 1980. "Firm Relocation and Washington, D.C. Processed. Site Selection in Suburban Municipalities." Journal of Meyer, J. R. 1981. "Report on Proposed Korean Spatial Study." Urban Economics, vol. 8, no. 1 (July), pp. 69-85. Urban and Regional Report 81-1. World Bank, Washing- Friedman, J. 1975. "Housing Location and the Supply of Local ton, D.C. Processed. Public Services." P-5421. Santa Monica, Calif.: Rand Cor- Mieszkowski, P., and M. Straszheim, eds. 1979. CurrentIssues poration. in Urban Economics. Baltimore, Md.: Johns Hopkins Uni- Hanushek, Eric A., and Byung Nak Song. 1978. "The Dynamics versity Press. of Postwar Industrial Location." Review of Economics and Mills, E. S. 1972. Urban Economics. Glenview, Ill.: Scott, Statistics, vol. 60, no. 4 (November), pp. 515-22. Foresman. Henderson, J. V. 1980. "A Framework for International Com- Murray, Michael P. 1982. "Here, There, Where? A Strategy for parisons of Systems of Cities." Urban and Regional Report Evaluating Industrial Relocation Policies in Korea." Urban 80-3. World Bank, Washington, D.C. Processed. Development Discussion Paper 6. World Bank, Washing- Hoover, E. M., and R.Vernon. 1959.AnatomyofaMetropolis. ton, D.C. Cambridge, Mass.: Harvard University Press. Muth, R. 1969. Cities and Housing. Chicago: University of Ingram, G. K. 1977. Residential Location and Urban Housing Chicago Press. Markets. Cambridge, Mass.: National Bureau of Economic Quigley, John. 1976. "Housing Demand in the Short Run: An Research. Analysis of Polytomous Choice." Explorations in Economic Kain, John F., and John M. Quigley. 1975. Housing Markets Research, no. 3. Cambridge, Mass.: National Bureau of Eco- and Racial Discrimination: A Microeconomic Analysis. nomic Research. Available from University Microfilms, Ann Cambridge, Mass.: National Bureau of Economic Research. Arbor, Mich. Kemper, P. 1973. "The Location Decisions of Manufacturing Renaud, Bertrand. 1981. National Urbanization Policy in De- Firms within the New York Metropolitan Area." Ph.D. dis- veloping Countries. New York: Oxford University Press. sertation. Yale University, New Haven, Conn. Processed. Schmenner, R. 1973. "City Taxes and Industry Location." Lee, K. S. 1981. "Intra-Urban Location of Manufacturing Em- Ph.D. dissertation. Yale University. New Haven, Conn. Pro- ployment in Colombia."JournalofUrbanEconomics, vol.9, cessed. no. 2 (March), pp. 222-41. . 1982. Making Business Location Decisions. Engle- . 1982a. "Determinants of Intra-urban Location of wood Cliffs, N.J.: Prentice-Hall. Manufacturing Employment: An Analysis of Survey Results Solow, R. 1972. "On Equilibrium Models of Urban Location." for Bogota, Colombia." Urban and Regional Report 81-3. In Michael Parkin, ed., Essays in Modern Economics. Lon- World Bank, Washington, D.C. Processed. don: Longman. - 1982b. "A Model of Intraurban Employment Location: Straszheim, M. R. 1974.AnEconometricAnalysisofthe Urban An Application to Bogota, Colombia." Journal of Urban Housing Market. Cambridge, Mass.: National Bureau of Economics, vol. 12, no. 3 (November), pp. 263-79. Economic Research. - . 1985. "An Evaluation of Decentralization Policies in Struyk, Raymond J., and Franklin J. James. 1975. Intrametro- Light of Changing Location Patterns of Employment in the politan Industrial Location. Lexington, Mass.: Lexington Seoul Region." Urban Development Discussion Paper 60. Books. Employment Location and Spatial Policies: Colombia and Korea 105 Tolley, George S., Philip E. Graves, and John L. Gardner, eds. Vol. 10, pt. 2 in D. R. Diamond and J. McLoughlin, eds., 1979. Urban Growth Policy in a Market Economy. New Progress in Planning. Elmsford, N.Y.: Pergamon. York: Academic Press. World Bank. 1979. World Development Report 1979. New Townroe, D. M. 1979. Employment Decentralization: Policy York: Oxford University Press. Instrmnents for Large Cities in Less Developed Countries. 9 Industrial Deconcentration Policy: Venezuela Benjamin Reif I n 1974 the Venezuelan government, conscious of gional disparities in wealth and well-being. Venezuela's problems caused by the concentration of activities in five areas of industrial deconcentration are shown in the north central area around the cities of Caracas, map 9-1. Maracay, and Valencia and the limited economic activity The main question is whether government instru- in lagging areas, decided to intervene in the spatial ments, including incentives and restrictions, have con- distribution of the country's manufacturing industry. A tributed to the deconcentration of industry. This series of presidential decrees provided a legal framework question brings up more specific questions. Has decon- for an industrial deconcentration policy. Evaluations of centration of industry taken place? Is deconcentration a the results show little agreement: a report by the Minis- result of the policy instruments? What other factors try of the Environment (Ministerio del Ambiente 1979a, have influenced industrial location, and what has been p. 81) claims that the behavior of the entrepreneur is their relative importance? The following sections at- changing, but a document prepared by the National tempt to answer these questions. Planning Office (Cordiplan 1979, pp. 7-9) indicates that the policy merely accentuated exist ing tendencies toward concentration of manufacturing activity in met- Has Deconcentration of Industry ropolitan Caracas and its surroundings. This study at- Taken Place? tempts to throw light on the debate. Venezuela's industrial deconcentration policy had The effects of industrial deconcentration policy are three aims: to prohibit the location of new manufac- expected to be felt at the regional level. The firm is the turing in the metropolitan region (Caracas and its sur- basic statistical unit, and in this study it is defined to be roundings); to induce industries that are considered identical to an establishment. Much of the data, such as hazardous to move to safer places and to encourage the number of employees and the production value, is nonhazardous firms already located in the congested tabulated by firm. On the basis of aggregate data for metropolitan region to move to designated development administrative regions, used to identify changes in the areas; and to attract new manufacturing plants to the distribution of establishments (table 9-1), it can hardly designated development areas. The main policy instru- be concluded that the spatial pattem of industry by ments for carrying out industrial deconcentration have administrative region has changed substantially. At been direct incentives (including financial and fiscal most, it can be said that between 1974 and 1976 the incentives such as grants, loans, and tax concessions), concentration of manufacturing in the Capital region indirect forms of assistance, such as the provision of increased by 5 percent at the expense of the other re- industrial parks and the improvement of transport and gions. communication facilities, and negative incentives, such It may be argued that the results, when presented as a as control over location. Incentives vary among zones in whole, may fail to illuminate the fact that individual accordance with the national goal of reducing interre- industrial subgroups may behave differently. Several 106 Industial Deconcenhrton Policy: Venezuela 107 Map 9-1. Industrial Deconcentration Areas in Venezuela < /> zCaribbean Sea D TRINIDAD J Moracaibo Cabimas CARASTOANGO 100 4 ) Barquisimeto° -1 a _ San Crist6bol Civdad Guayona Area A I .jY ) .Area B C' .t Area BMX , .Area C BRAZIL| . ,Area D \' -tBRZ - -- International Boundaries< > -go B f 0 ~~~~~~~~~~~~~~~~~~200 KILOMETERS 70'-0/' \~^'~2J'0 600 manufacturing sectors have therefore been selected for industries and with local input and product markets, as analysis. (A study of all the sectors is beyond the scope of indicated by the input-output table of the Venezuelan this research.) economy, to include the possible effect of spatial policies Of the twenty-seven three-digit ISIC (Internation- on the location of transport-oriented industries. The al Standard Industrial Classification) manufacturing levels of effective protection for the industries had to be codes, sectors having between 140 and 300 establish- diverse, as did their intensities of capital and labor. (The ments-the middle range-were chosen for study. mean electricity consumption value, in kilowatt-hours Eleven sectors were initially selected, after excluding (1) per thousand hours, was used as a proxy to reflect any industrial sectors with many establishments that serve tendency toward deconcentration of labor-intensive in- local markets and are little affected by regional policy dustries.) Six three-digit ISIC codes were chosen for (for example, bakeries) and (2) sectors that have such a further study: textiles (321), leather products (323), small number of plants (probably serving national or plastic products (356), nonelectrical machinery and international markets, as is the case with oil refineries) equipment (382), electrical machinery and equipment that studying them' would yield only limited insights (383), and transport equipment (384). into locational behavior. Selected industries also had to Venezuelan manufacturing is concentrated in an axis experience a high rate of growth. The industries chosen that extends over parts of the Capital and Central re- had to have strong linkages with other manufacturing gions. One way to ascertain whether deconcentration of 108 Berammin Reif Table 9-1. Shares of Manufacturing Establishnents, By comparing the 1974 and 1976 industrial directo- by Administrative Region ries, changes that occurred in 1975 and 1976 were iden- (percent) tified. Similarly, by comparing the 1976 and 1978 direc- Region 1971 1974 1976 1978 tories, changes during 1977 and 1978 were identified. Capital 51 50 55 54 Once the directories were matched, factors could be Central 16 17 16 15 examined for each period-for example, which firms Centro-Occidental 8 7 7 7 were new (founded during the period); which closed Zuliana 8 9 7 7 down (went out of business); which were stationary Andes 7 8 7 8 (remained in the same location); and which moved to Sur 1 0 0 0 another location or from another region. The results of Nor-Oriental 7 6 5 6 the matching (table 9-3) show the share of new estab- Guayana 2 3 3 3 lishments in the Venezuelan industrial axis (areas A, B, Memorandum: and BM on map 9-1). Number of Each of the six industries shows a similar trend establishments 6,401 7,554 9,538 10,478 toward deconcentration. Some of this may be a conse- Sources: Cordiplan (1973); Oficina Central de Estadistica e Infor- quence of the big increases in income and demand in matica (1974, 1976, 1978). 1974-76 as a result of oil price increases. The effects of these increases were perceived mainly in Caracas and led to an unusually high concentration of new firms. Never- Table 9-2. Shares of Manufacturing Establishments theless, as table 9-3 shows, for leather, nonelectrical in the Capital and Central Regions Combined machinery, and electrical machinery the share of the (percent) industrial axis in the national total decreased by more Industry 1971 1974 1976 1978 than 33 percent by the second period. It thus appears Textiles 84 88 88 88 that between periods these three industrial subgroups Leather 76 71 77 75 experienced deconcentration. Declines in the shares of Plastics 97 90 90 89 the other manufacturing groups-textiles, plastics, and Nonelectrical transport equipment-are mainly a result of the initial machinery 69 71 77 75 distortions created by the big increase in demand and Electrical therefore are not necessarily evidence of decentraliza- machinery 91 88 90 87 tion. Transport The absence of disaggregated data before 1974 places equipment 71 72 75 72 limits on the conclusions. When an industrial subsector Sources: Same as for table 9-1. shows a tendency to deconcentrate, the tendency may have started in 1974 or may have its origins in an earlier period. This uncertainty makes it difficult to relate the manufacturing has taken place is to combine the shares of the Capital and Central regions, as in table 9-2. No major changes took place during the period of the study Table 9-3. Share of New Manufacturing except in the plastics and nonelectrical machinery in- Establishments in the Industrial Axis dustries, which show a relatively higher share for 1971 (percent) and 1976, respectively; that is, no significant evidence of Industry 1974-76 1976-78 Change concentration or deconcentration can be found. Because the results are not given for the officially Textiles 89 (82) 81(26) - 8 defined areas of industrial deconcentration, a proper Leather 81 (52) 47 (17) -34 evaluation of the effects of the policy instruments is Plastics 92 (98) 74 (27) -18 difficult. Furthermore, aggregate figures like those used Nonelectrical here do not permit assessment of the spatial changes in machinery 87 (87) 51(33) -36 the manufacturing industry. The data are for the net Electrical machinery 88 (61) 33 (12) - 55 number of establishments that existed in a particular Transport year and do not show the actual numbers of new firms equipment 76 (74) 59 (34) -17 established in an area, of firms that went out of business, Note: Number of new firms is in parentheses. or of plants that changed locations. Attempts were made Source: Author's estimates on the basis of the matched directories to find data that overcome these limitations. and Oficina Central de Estadfstica e Informatica (1978). Industrial Deconcentration Policy: Venezuela 109 pattern of industrial deconcentration to govemment Figure 9-1. Actual and Extrapolated Share policies. But the important issue is whether deconcen- of Credits, Three Regions tration of manufacturing took place over the two periods for which there is evidence for three of the six subsectors 70 analyzed. 63.6 Is Deconcentration a Result of Policy? 60 - Periphery 55.9 Owing to difficulties in measuring the effects of re- 50 - gional policies (variables may be affected by migration 48.2 and long-term structural changes as well as by regional X policy), it is not possible to employ conventional statis- G tical methods for rejecting or accepting hypotheses. 4 Instead, we will assess alternative hypotheses on the X basis of indirect evidence. Additional evidence from econometric models is discussed in subsequent sections. °1 30Central 24.7238 Cnrl 2. Financial Incentives 2020 22.3 If industrialists took advantage of government- 1 sponsored financial incentives in developing areas, this C 12.7 should be reflected in the number of credit applications 10 in these areas. Available data are for number of credits awarded, not for number of applications; it is assumed 0 that the proportion of applications approved does not 1973-74 1975-76 1977-78 vary by region. The number of credits granted to a (Period I) (Period II) region was thus used as a proxy for the number of (PActual applications. Similarly, the data did not differentiate ---- Extrapolated between credits for new firms and for expansions; it was assumed that the proportion of expansions to new firms foreign ownership, is able to earn through industrial did not vary by region. investment, In relation to the average profit margins of Figure 9-1 shows the actual and extrapolated shares of Venezuelan manufacturing industry, the saving owing credits for the Capital, Central, and peripheral regions. to fiscal incentives may be insignificant in many cases. The results suggest that the regional trends existed before the introduction of financial incentives in March 1976 and continued over the next two years, and that the Negative Incentives introduction of the incentives may have had little addi- As early as January 1975 a decree was issued that tional effect on the spatial distribution of manufac- prohibited new establishments from locating in the met- turing. ropolitan area (Caracas and its surroundings) unless Fiscal Incentives they qualified under special conditions. Furthermore, the government drew up three lists of industries, iden- A study by the Ministry of the Environment (Ministe- tified here as groups 1, 2, and 3. A firm's right to rio del Ambiente 1979b, p. 33) estimates that the savings continue to operate or to locate in area A depended on from tax holidays represent only 1.5 to 2 percent of sales the list to which it was assigned. It is assumed that the value. World Bank data indicate that-subject to caveats lists, although approved in 1976, had actually been in regarding statistical reliability-industrial profit rates effect since January 1975, when the prohibition on are very high in Venezuela. When petroleum refining locating in area A first took effect. The lag constituted an and the state-enterprise-dominated basic metal indus- informal grace period to reduce hardships on firms tries are removed from the total the average industrial whose location was already planned. profit rate, net of depreciation and taxes, in the 1970s rises from 19 to 28 percent, a figure that more accurately New FiYrms. New firms in group 1 are the only ones reflects what the private sector, of both national and allowed to locate in the metropolitan area. All others are 110 Benjamin Reif considered "transgressors." The table shows that there period. The framed cells represent the expected impact was a lower percentage of transgressors in 1975-76 than of deconcentration policy. In spite of the reduced num- in 1977-78, although the absolute number of transgres- ber of cases, the general trend seems to be one of con- sors was higher.' centration. The evidence indicates that negative incen- tives were not applied effectively during the study Number of new Number of Percentage ofd Period firms in area A transgressors transgressors period. I (1975-76) 290 127 43.8 11(1977-78) 64 33 51.6 Preliminary Conclusions. There seems to be little evidence that changes in the spatial pattern of industry An explanation for the lower percentage in 1975i76 may are related to the application of government instru- be that government officials were stricter during that ments. The nearly stable trend in regional shares of period. There are no disaggregated data before 1975, and credits does not support the hypothesis that financial it is not possible to know what was happening before the incentives influenced the location of new firms. In- location controls were imposed. It may be that the per- formation on the actual application of fiscal incentives is centage of firms identified with groups 1 and 2 that generally not available, but the impact of such incen- located in the metropolitan area was higher before the tives is likely to be insignificant compared with the controls were imposed. Or, the controls may not have profits of the manufacturing industry, and it is probable been implemented effectively in 1977-78. that they have had little effect on location. Negative incentives do not seem to have been applied effectively Moves of Existing Establishments. Figure 9-2 shows during the study period. an origin-destination matrix of interarea moves for each Other Factors Figure 9-2. Origin-Destination Matrix of Interarea The evidence suggests that some manufacturing in- Movers, 1975-76 and 1977-78 dustries had a tendency to deconcentrate and that gov- emment policies had little effect on industrial deconcen- 1975-76 tration. This leaves deconcentration unexplained and leads toward investigation of other factors that influence To area Moved Moved Net industrial location. From area A B IBM C D out in change Changes in the Venezuelan economy during 1974, _________ - - a - - when the national budget increased threefold, probably A 7 4 11 12 + 1 had two significant effects on manufacturing. First, B 1 1 7 + 6 many new manufacturing firms appeared that were pri- marily concerned with satisfying the sudden increase in BM 12 12 5 - 7 demand and that benefited from the high profits. Dur- C 0 0 o ing this period most entrepreneurs had a short-term - -- - - - commercial bent rather than a long-term interest in manufacturing. As a consequence, most of the firms established during 1975-76 chose sites without much 1977-78 concern for cost or other locational factors. Second, the steep increase in the number of manufacturing estab- To area Moved Moved Net lishments aggravated the diseconomies of the physical From area A B BM C D out in change and social infrastructure. This started in Caracas and _ - - slowly spread to many of the surrounding cities. New A 2 2 4 5 + 1 industries that located in places where the level of ser- B 2 -2 2 0 vices was inadequate suffered along with plants already B 2 2~~~~~~ 2 0there. BM 3 3 2 -1 By the second period (1977-78) the economy had D 0 0 0 reached equilibrium in the supply of and demand for … -0 1 - -v national manufactured products. By then most entre- preneurs were not only aware of the problems of locating in the congested industrial axis but were also aware that Note. The framed cells represent the expected impact of the policy. Source: Author'sestimatesonthebasisofthematcheddirectoriesand the bonanza, or high-profit, perod was over. These on Oficina Central de Estadistica e lnformatica (1968). entrepreneurs were therefore more concemed with lo- Industrial Deconcentration Policy: Venezuela 111 cational factors than were entrepreneurs during period diseconomies of being in the industrial axis, we might 1, and they probably planned more carefully the selec- expect highly protected industries to concentrate and tion of their plant sites, looking for places with lower less protected industries to deconcentrate. The six wages, a satisfactory water supply, a good social infra- manufacturing groups considered here behaved as ex- structure, and low transport costs. The search for those pected: textiles, plastics, and transport equipment are characteristics may have led some industrialists to con- highly protected and are concentrated; leather, non- sider settlements outside the industrial axis, and conse- electrical machinery, and electrical machinery face a low quently deconcentration occurred. Deconcentration level of protection and are deconcentrated.2 probably began before the economic boom of 1974, but One could argue that industries which concentrated the sudden change in the economy at that time may have did so not because of government protection but for temporarily slowed the process. other reasons and that the pattern of association de- These arguments may explain why some manufactur- scribed above is spurious. This is possible, and we do not ing sectors deconcentrated, but what about those which want to present the results shown here as conclusive. did not? A clue may be found in World Bank research Attempts to find other causes were made. For example, which indicates that the salient feature of the Vene- we tested capital or labor intensity, the product cycle zuelan industrial environment in the 1970s was the high theory, and other factors, but none were associated with degree of protection from import competition accorded concentration. by government policies. This suggests that highly pro- The next tasks were to determine whether less pro- tected industries were attracted to Caracas. For example, tected industries were more concerned with factors of small, highly protected industries would probably con- location, to identify these factors, and to evaluate the sider closeness to the government's decisionmaking extent to which government financial incentives have center to be crucial; exonerations, licenses, and other been considered as a locational factor by industrialists. benefits received on a case-by-case basis might outweigh For this purpose a discrete choice logit model was used the disadvantages of locating in the industrial axis. Mur- that represents an entrepreneur who considers the attri- ray (1982) states that "access to government is one input butes of both the firm and the city in deciding where to in many production processes." Renaud (chapter 5, locate a manufacturing plant. The model relates the above), in a discussion of the unintended biases of variables of the cities to the locational choice of the national economic policies, stresses that the manage- entrepreneur. The choice is made from a known set of ment practices of the central government and its regula- alternative cities. Modelers are faced with the task of tion of economic activities require location of these explicitly defining this location choice set. In one re- activities close to the capital and contribute to the urban spect this is virtually a hopeless task: Venezuela consists vortex. A complementary explanation would be that the of many settlements of varying size and quality, and the savings that could have been obtained through decon- subset actively considered by any given entrepreneur is centration were low compared with the large profits probably known to the industrialist alone. Central place made possible by the high level of protection. theory, however, argues that a natural hierarchy, or If this argument holds, we would expect to find sup- system of settlements, exists within a country. If such a porting evidence, for example: hierarchy exists, it is possible to identify the most impor- * Concentration of highly protected manufacturing tant settlements for manufacturing activity. (These set- in the industrial axis and deconcentration of less tlements presumably constitute a large sample of possi- protected manufacturing firms ble locations for those manufacturing entrepreneurs Groteted manuacturnwin focatirm on the part of less who desire a regional or national market for their prod- * Greater concern with location on the part of less ucts.) For such a hierarchy the modeling technique protected industries; a higher sensitivity of second- employed here is useful. The logit model assumes one period industrialists to locational factors, owing underlying utility function and looks at relative choice both to their awareness of the physical limitations frequencies as sample draws. in cities of the industrial axis and to a return to more Because of computational limitations-it is unlikely realistic profit levels that existing computer programs could adequately * Relatively little weight attached to government in- handle more than twenty potential choices of location- centives as a factor in location. a limited number of alternative locations was selected. The rest of this section will address these issues. For this purpose, several criteria were applied. Concentration and protection * The choice set should be large enough to adequately represent the variables to be analyzed, that is, to If the benefits of being close to government agencies have a wide range of values for the independent that grant exonerations and other benefits outweigh the variables. 112 Benjamin Reif *The choice set should include a mix of core and that is, the net effect of all attributes of an alternative peripheral cities as well as settlements with and which is not measured by the other variables. A dummy without manufacturing. The manufacturing need variable was assigned to Caracas because it is the loca- not be new. tion chosen by about 66 percent of new manufacturing * The choice set should represent all areas of decon- industries. centration as well as the special growth centers The variable FI represents the financial incentives of identified by the policy of industrial deconcentra- the industrial deconcentration policy. It incorporates a tion. number of elements, including interest rates, grace and * The choice set should have geographic balance; it loan periods, and uses of money, whether for land and should include cities located in the east, west, and construction, machinery and equipment, or working south of the country as well as in the traditional capital. north-central area. Total wage (TWAGE) represents the interaction be- tween the prevailing wages of the city and the number of With the use of this selection procedure, twenty cities employees in the firm-that is, the total payroll that were chosen. each firm would have in each of the twenty cities. The interaction reflects the hypothesis that firms with large Specification of the Logit Model payrolls are proportionately much more sensitive to labor costs than are smaller ones. Furthermore, it may Manyindeendet vaiabls ma affct lcatinal be hypothesized that firms tend to locate where wages choice. To reduce them to a manageable set, we ex- b . . f amined the statistical correlation among a large number are lower and hence a negative coefficient is expected. amlna tn stastlcl corelaon aong lare nuDer The agglomeration potential associated with local in- of variables for which we had data. High correlations Th gglmrtopteiaascaedwhlclin betweengroupsof variablesforwhich were had odat horrex s puts, LIP, was designed to reflect a firm's accessibility to between accessibility to local inputs and accessibility to both local inputs and product markets in each city. It is a bwna accsibili. to l . exainpts and ac cesiablit to composite variable that reflects not only industrial link- fin ftt-al cos. r exami.n estion te viricable i ages but also transport costs. It is assumed that as the terms of statistical correlation, existing empirical work distance to a particular source of inputs increases, the (see, for example, Carlton 1979), reliability of the attractiveness of that source decreases. source, and, especially, the economic theory of location, The variable TEC reflects the hypothesis that entre- the independent variables were selected. They are as follows: preneurs look for places where technicians and related professionals are available. This should be especially true Ul 1 in the Caracas alternative, 0 otherwise for firms with a high level and proportion of skilled F! Percentage savings in present value terms workers. A positive coefficient is expected. owing to financial incentives (log) The variable STR reflects the hypothesis that entre- TWAGE Wage times number of employees in the preneurs tend to locate their plants away from areas of firm (log) high labor unrest. It is measured in terms of average LIP Agglomeration potential associated with working hours lost per thousand workers per year be- local inputs (log) cause of unexpected stoppages. A negative coefficient is expected. TEC Proportion of the city's professionals, tech- The variable PHN reflects the effect of telephone ser- nicians, and related occupations to its vice on an industry's desire to locate in cities with population (log) telecommunication facilities. A positive coefficient is STR Average number of working hours lost per expected. thousand workers per year because of unex- The variable DOC measures the number of inhabi- pected stoppages (log) tants per doctor. It may be hypothesized that entre- PHN Number of telephone lines (installed capac- preneurs are more attracted to cities with better health ity) per thousand inhabitants (log) facilities, and it is assumed that a lower number of DOC Number of inhabitants per doctor (log) inhabitants per doctor implies better health services. A FIP Agglomeration potential associated with negative coefficient is expected. foreign inputs (log) The agglomeration potential associated with foreign inputs, FIP, was designed to reflect the accessibility to WPR Percentage of the city's population serviced foreign inputs in each city. It is a composite variable that with water (log) reflects dependence on foreign raw materials and com- The variable U! is a constant term in the utility func- ponents and the transport costs to international ports. tion that measures the so-called pure alternative effect, The hypothesis is that the higher the proportion of Industrial Deconcentration Policy: Venezuela 113 Table 9-4. Parameter Estimates for FYve Model Specifications for 1975-76 Model specification Variable 1 2 3 4 5 UI 0.35 - 0.95 - 2.85 -3.04 -1.06 (0.87) (1.50) (2.71) (2.88) (1.75) F! - - - - - TWAGE -1.32 - 0.91 -1.42 -1.45 -1.47 (2.90) (1.93) (3.16) (3.21) (3.18) LIP 1.12 1.16 1.36 1.29 1.09 (8.18) (8.06) (7.48) (6.91) (7.19) TEC 1.90 3.20 3.26 3.41 2.71 (3.01) (4.20) (4.25) (4.48) (3.90) STR - 0.19 -0.16 - 0.49 - 0.51 - 0.31 (1.99) (1.52) (3.69) (3.68) (2.77) PHN - 1.24 - 0.76 1.15 (3.21) (1.94) (3.16) DOC - - -2.37 - 0.82 - (3.52) (2.46) FIP - - 0.50 - - - (3.74) WPR 1.39 - 4.12 3.78 1.59 (1.83) (3.28) (2.83) (1.73) Sum of squared residuals 12,350 19,890 43,820 40,660 15,300 Degrees of freedom 7,157 7,156 7,156 7,155 7,156 Percentage correctly predicted 70.29 70.29 70.29 70.29 70.29 Likelihood ratio index 0.5637 0.5742 0.5696 0.5714 0.5686 Likelihood ratio statistic 1,273 1,297 1,287 1,291 1,284 -Not applicable. Note: Number of firms in the sample, 377. Numbers in parentheses are t-statistics. imported components for local assembly, the closer the variable FIP. Models 3 and 4 are dropped because of high assembly plant will be to international ports. A positive variation in the parameter values of the variable DOC. coefficient is expected. We disregard model 5 because it has one more variable The variable WPR reflects the hypothesis that indus- than model 1 and adds no additional explanatory power. triestendtolocatewherewaterisavailable.Thisvariable (Both models have a likelihood ratio index of 0.56.) may be considered a proxy for the availability of infra- Furthermore, the unexplained residuals are higher in structure services. model 5. (The sum of squared residuals is 15,300 in model 5 and 12,350 in model 1.) Hence, of the five specifications we select the first; indeed, specification 1 was selected out of all those analyzed as the model to be For any applied model of locational choice, the selec- used in this research. The likelihood ratio index of 0.56 tion of an appropriate specification must include experi- for this model indicates that its overall goodness of fit is mentation with the data. It is, of course, important to satisfactory. avoid mining the data. To guard against this danger, all To analyze the data for the second period (1977-78), experimentation was done with the subsample of data variable FI (financial incentives) was added to specifica- for the period before the introduction of the financial tion 1, and the parameters were estimated. The results incentives package. As will be seen, this model fits the appear in table 9-5, which allows us to compare the second subsample equally well. results from the model with data from the two periods Several distinct logit model specifications were esti- 1975-76 and 1977-78. The model seems to be structur- mated with data for the first period. These models cor- ally stable over time. The signs of the coefficients (except respond to different combinations of the variables UI) are the same for both periods. Furthermore, the selected earlier. The details of five of the model specifica- values of the coefficients are similar except for WPR tions estimated are given in table 9-4. Of the five models, (availability of water or infrastructure services), which we reject model 2 because of the unexpected sign on the shows a significant increase in the second period. 114 Benjamin Reif Table 9-5. Parameter Estimates for the Sample work stoppages, STR, is not large but is statistically of Six Industry Subgroups, 1975-76 and 1977-78 significant in both periods, which implies that entrepre- Variable 1975-76 1977-78 neurs are concerned with locating their plants away from cities with high labor unrest. The coefficient for Ul 0.35 -0.06 the variable measuring water availability, WPR, in- (0.87) (0.08) creases more than three times in the second period and Fl (0.31) becomes even more statistically significant. A close look 7WAGE -1.32 -1.69 at variable WPR in table 9-5 indicates two issues. First, (2.90) (3.86) manufacturing firms seem to be attracted to places LIP 1.12 0.96 where water (or infrastructure services) is available. (8.18) (5.26) Second, it is likely that second-period entrepreneurs TEC 1.90 1.41 were more sensitive to availability of water (or infra- (3.01) (1.50) structure services). STR -0.19 -0.21 On the basis of these results it seems reasonable to (1.99) (1.91) conclude the following. WPR 1.39 4.79 (1.83) (3.45) * Wages are an important factor in explaining the Sum of squared residuals 12,350 1,914 location of industry. Degrees of freedom 7,157 2,235 * The evidence provides little support for the proposi- Percentage correctly predicted 70.29 52.54 tion that financial incentives have exerted an effect Likelihood ratio index 0.5637 0.3683 on the location of manufacturing. Likelibood ratio statistic 1,273 260.4 * Accessibility to both local inputs and product mar- Memorandum: Number kets exerts a large influence on the locational deci- of firms in the sample 377 118 sion of a manufacturing firm. - Not applicable. * The presence of technical expertise is probably an Note: Numbers in parentheses are t-statistics. important locational factor. Analysis of the Coefficients * The more labor unrest a city has, the less likely it is to attract new manufacturing. The dummy coefficient for Caracas, Ul, is subject to a * The availability of water (or infrastructure services) wide confidence interval in both periods, as is shown in exerts a major influence on the location of new table 9-5. The net effect of all attributes of Caracas which manufacturing. It is likely that this variable was are not measured by the other variables may be the most more important during the second period than dur- important factor in attracting new industry, but the data ing the first period. simply do not allow us to determine this effect with much precision. Indeed, the opposite could be the case; Results, by Industry Subgroups Caracas may exert a negative effect. (Some evidence of this is seen in table 9-4.) In those cases in which we have So far the results have dealt with all six industrial controlled for more variables, a negative coefficient for subgroups together. The next question is whether the Caracas is obtained. results apply to each subgroup. One might expect the The coefficient of the financial incentives variable (FI) importance of the attributes to vary among manufactur- is very small and statistically insignificant. The failure of ing groups. financial incentives to show up as an important in- The number of establishments (by industry subgroup fluence on location is consistent with our previous and by year of foundation) to be used in our econometric finding. Wage levels, especially when they interact with analysis is less than the total number of new establish- the number of employees, TWAGE, exert a large in- ments because firms that locate outside the choice set fluence on the location of industry. The coefficients are (the twenty cities selected) are not considered. Because statistically significant in both periods. The accessibility electrical machinery had only nine establishments in the to sources of local inputs and to product markets, LIP, second period, and it is unlikely that we can infer any enters significantly in both periods; the coefficients are meaningful conclusions from the analysis of such a statistically significant. Having a pool of technical exper- small number of observations, results for electrical tise in the city, TEC, as measured by the proportion of machinery for the second period are omitted. the city's professionals, technicians, and related occupa- The results, stratified by manufacturing groups, are tions to its population, seems to be an important factor presented in tables 9-6 to 9-8. For plastics, the max- in location. The coefficient of the variable measuring imum likelihood routine had difficulty converging when Industrial Deconcentration Policy: Venezuela 115 Table 9-6. Parameter Estimates for the Textile and Transport Equipment Industries, 1975-76 and 1977-78 Transport Textiles equipment Variable 1975-76 1977-78 1975-76 1977-78 UI 0.05 -0.09 -2.38 -0.33 (0.02) (0.02) (1.78) (0.16) Fl - 0.04 - 0.04 (0.38) (0.25) 7WA GE -2.66 -2.54 -3.31 0.14 (0.88) (0.79) (1.78) (0.13) LIP 1.42 1.36 2.00 1.23 (1.79) (1.54) (3.31) (2.36) TEC 4.69 2.21 3.54 - 0.16 (1.78) (0.51) (1.85) (0.07) STR - 0.22 -0.09 - 0.55 -0.47 (0.80) (0.31) (2.12) (1.68) WPR 9.73 20.93 1.22 6.88 (1.37) (1.66) (0.72) (1.52) Sum of squared residuals 795.7 267.7 929.4 465.1 Degrees of freedom 1,286 430 1,153 468 Percentage correctly predicted 86.76 78.26 44.26 52.0 Likelihood ratio index 0.7685 0.67 0.3478 0.3793 Likelihood ratio statistic 313.1 92.32 127.1 56.81 Memorandum: Number of firms in the sample 68 23 61 25 - Not applicable. Note: Numbers in parentheses are t-statistics. the WPR variable was included in the model for 1977- larger during period II; that is, it is hypothesized that 78. Hence, in table 9-8 we present the results for plastics during period II entrepreneurs of less protected indus- for period I only. To investigate the possible effects of the tries searched more carefully for places that offered independent variables on locational decisions in the reduced costs and increased revenues. plastics industry, several other specifications without The pattern of association between the protection WPR were analyzed, and the results are presented in level and the expected behavior of wages shows that table 9-9. leather and nonelectrical machinery, with low levels of protection, have larger wage coefficients (TWAGE ) dur- Wages (TWAGE). We indicated earlier that, for the ing period II, and textiles and transport equipment, unstratified sample of industries, wages seem to exert a which are highly protected, do not, so they behave as large influence on location. Under stratification the expected. (Plastics and electrical machinery are not con- wage coefficient for all six industries is always subject to sidered because of the limitations indicated earlier.) a wide confidence interval (see tables 9-6 to 9-8). Except Thus, less protected industries are more sensitive to for the plastics industry, the expected negative sign wages during period II, in accordance with our hypoth- predominates. Wages seem to be very important for the esis. Even if the plastics and electrical machinery indus- leather, nonelectrical machinery, and textiles indus- tries are considered, these conclusions hold. This tries. Furthermore, for leather and nonelectrical ma- finding lends additional support to the hypothesis that chinery the coefficients in period II are larger, which protection for manufacturing may suppress the entre- may imply an increased sensitivity to this variable. preneur's interest in locating in low-wage areas. The hypothesis proposed here is that less protected industries are more concerned with locational factors Access to Local Inputs and Product Markets and that the economic bonanza of the first period may (LIP). The coefficients for the variable which measures have masked the effects of locational factors. We expect access to sources of local inputs and product markets coefficient values for the less protected industries to be (LIP) are usually important for all six manufacturing 116 Benjamin Reif Table 9-7. Parameter Estimates for the Leather and Nonelectrical Machinery Industries, 1975-76 and 1977-78 Nonelectrical Leather machinery Variable 1975-76 1977-78 1975-76 1977-78 UW 5.26 -2.26 1.26 -0.20 (2.02) (0.67) (1.36) (0.13) Fl - -0.16 - 0.15 (0.99) (0.97) 7WAGE - 2.66 -12.23 -0.86 - 2.26 (0.73) (1.40) (0.49) (0.94) LIP 0.66 1.15 1.03 0.96 (0.95) (1.10) (2.89) (2.36) TEC 1.01 7.77 0.90 2.82 (0.30) (1.32) (0.43) (0.76) STR -0.46 - 0.48 - 0.26 0.10 (1.57) (0.54) (0.80) (0.26) WPR 10.36 9.71 1.15 4.09 (2.32) (1.76) (0.61) (1.22) Sum of squared residuals 803.1 190.6 1,364 345.2 Degrees of freedom 887 259 1,343 449 Percentage correctly predicted 87.23 57.14 73.24 29.17 Likelihood ratio index 0.7891 0.47 0.5831 0.2118 Likelihood ratio statistic 222.2 39.50 248.0 30.46 Memorandum: Number of firms in the sample 47 14 71 24 -Not applicable. Note: Numbers in parentheses are t-statistics. Table 9-8. Parameter Estimates for the Plastics groups (see tables 9-6 to 9-8). Coefficients for transport and Electrical Machinery Industries, 1975-76 equipment, nonelectrical machinery, and electrical Electrical machinery are statistically significant, as is also true for Variable Plastics machinerg plastics, especially in the four model specifications that UI 1.87 -0.59 appear in table 9-9. (2.22) (0.63) FI - - Technical Expertise (TEC). When results are strat- 7WAGE 1.09 -0.42 ified by industry subgroup, technical expertise exerts (0.74) (0.17) influence on each of the six manufacturing sectors. The LIP 0.45 6 (3.64) coefficients have wide confidence intervals (primarily TEC (1.51 1.20 during period I) except for textiles and transport equip- (1.15) (0.54) ment. Table 9-6 shows that textiles and transport equip- STR _-0.11 - 0.27 ment behave unexpectedly by lowering their coefficients (0.51) (0.72) during period II. Table 9-7 indicates that leather and WPR -0.93 0.42 nonelectrical machinery have larger coefficients in (0.60) (0.21) period II, in accordance with the expected behavior. Sum of squared residuals 1,727 764.8 The pattern of association between the level of protec- Degrees of freedom 1,495 963 tion and the expected behavior for technical expertise Percentage correctly predicted 74.68 52.94 provides additional evidence in support of the hypothesis Likelihood ratio index 0.5889 0.4428 that during period II less protected industries were more Likelihood ratio statistic 278.7 135.3 concemed with locational factors-that is, the values for Memorandum: Number the coefficients of the less protected industries during of firms in the sample 79 51 the second period are larger than the values for period 1. - Not applicable. Leather and nonelectrical machinery, which have low Note: Numbers in parentheses are t-statistics. levels of protection, have larger TEC coefficients; textiles Industrial Deconcentration Policy: Venezuela 117 Table 9-9. Parameter Estimates for Four Model Specifications for the Plastics Industry, 1977-78 Model specification Variable A B C D UI 67.32 69.26 70.71 67.08 (2.74) (2.65) (2.65) (2.66) Fl 7.30 8.21 7.11 7.52 (2.78) (2.76) (2.80) (2.78) TWAGE 1.57 3.07 -1.32 0.03 (0.93) (1.45) (0.29) (0.00) LIP 2.70 2.57 3.18 3.40 (2.52) (2.27) (2.31) (1.78) TEC 3.70 9.94 2.63 7.81 (1.07) (2.04) (0.68) (1.55) STR -1.33 -1.57 -1.32 -1.73 (1.81) (1.90) (1.64) (1.51) PHN - 4.75 - 6.38 (1.39) (1.15) DOC - - 5.70 7.70 (0.77) (1.06) Sum of squared residuals 168.1 95.92 159.8 87.13 Degrees of freedom 431 430 430 429 Percentage correctly predicted 60.87 52.17 60.87 52.17 Likelihood ratio index 0.5736 0.6055 0.5793 0.6150 Likelihood ratio statistic 79.04 83.44 79.83 84.75 Memorandum: Number of firms in the sample 23 23 23 23 -Not applicable. Note: Numbers in parentheses are t-statistics. and transport equipment, which are highly protected, do other four manufacturing industries. Furthermore, not, so they behave as expected. An explanation as to why with the exception of the leather industry, for which the some subgroups show a lower coefficient for technical coefficient in period 1I remains stable, the parameters expertise in period 1I may be that it is cheaper for the for the other three subgroups increase dramatically dur- firm to pay whatever amount is required to relocate its ing 1977-78. Again we may argue that this increase may technicians with it than to locate in places where tech- be a result of heightened awareness among second- nical expertise is available but where other locational period entrepreneurs. The reason that the coefficient factors are lacking. The additional cost is more easily remains unchanged during period II for leather manu- absorbed by high-profit firms, which usually are highly facturing can be found in the Isic 323 code, leather and protected. leather products, which includes tanning and leather finishing as well as fur dressing and dyeing. According to Labor Unrest (STR). The coefficient for labor unrest Fair and Geyer (1954, p. 856), high-quality water is is usually small and is subject to a wide confidence crucial to production for this subgroup, and it is unlikely interval, except in transport equipment, where it is sta- that first-period entrepreneurs would have overlooked tistically significant. An explanation for the high signifi- its importance. Hence, water availability was seen as cance (although low value) in this sector may be that it equally important during both periods by leather indus- includes motor vehicle assembly and the specialized trialists. manufacture of motor vehicle parts. Because workers in these activities are usually well organized, the transport Caracas Dummy (UI). The net effect of all attributes equipment sector is prone to conflicts. Locating plants of Caracas not measured by the other variables is away from areas of high labor unrest may be a measure reflected in the dummy variable, UI. Among the attri- to reduce the probability of unexpected stoppages. butes not captured are diseconomies related to conges- tion, land cost, and location control. Other factors, Water Availability (WPR). Plastics and electrical however, such as access to the bureaucracy, work in machinery seem to be little affected by water availability, favor of Caracas. but the coefficient for this variable is quite large for the The contribution of Caracas to the utility functions of 118 Benjamin Reif less protected firms is expected to decline with time machine cost. Usually, a different mold is needed for because of increasing diseconomies. For highly pro- each product. tected industries the disadvantages of Caracas also in- The relative ease of setting up a plastics factory con- crease with time, but this effect may be offset by the tributed to the growth of this sector. During period I, rising fear that protection will be removed (a likelihood seventy-nine new establishments were started, fifty-nine for textiles today), or by the search for additional ben- of them in Caracas. By period II private financial institu- efits, such as those related to local content regulation. tions had apparently become more aware of the risks Both factors make Caracas attractive. Hence, large nega- involved in extending loans to this sector because of the tive effects in period II cannot be expected for highly proliferation of plastics factories. The low flexibility of protected industries, since neamess to govemment is an plastics machinery may have contributed to this percep- advantage that counteracts the disadvantages of locating tion: in case of bankruptcy, a bank would have to take in Caracas. The pattern of association between the level over the plant and sell it, and a prospective buyer would of govemmental protection and the expected behavior have to be interested in producing products similar to for the Caracas variable shows that leather and non- the existing line or would have to make additional costly electrical machinery, with low levels of protection, have investments in molds. That the earlier owner went bank- a lower negative coefficient (U); textiles and transport rupt because of high competition lent the operation an equipment, which are highly protected, do not, so they added sense of risk. Consequently, during period 11 pri- behave as expected. vate financial organizations were probably reluctant to It may be argued that Caracas was not always per- give loans for manufacture of plastic products. Under ceived by entrepreneurs as experiencing diseconomies. these conditions an industrialist in plastics might well The larger negative effects observed in period II probably find it advantageous to move to the designated areas to represent a return to conditions similar to those that become eligible for govemment loans. Some industrial- existed before the distortions of the sudden economic ists therefore reacted favorably to the financial incen- changes of 1974. Unfortunately, the effects of location tives. It is likely that some of these entrepreneurs control cannot be separated from those of other disecon- selected locations in the area of deconcentration closest omies such as congestion and land cost. to Caracas (BM). By doing so, they remained close to the industrial axis and did not contribute to deconcentra- Financial Incentives (F1). In tables 9-6 and 9-7 the tion. coefficients on the variable for financial incentives are sometimes negative, always small, and usually statisti- cally insignificant. Until adequate data are available con- Conclusions clusions cannot be drawn for the electrical machinery and plastics industries. Only nine new electrical firms The hypothesis that less protected industries are more located within the twenty cities, too small a number for concerned with locational factors is supported by the statistical tests. For plastics we recall that the maximum evidence. More generally, this model attempts to iden- likelihood routine had difficulty converging with the tify the factors that affect the locational decisions of WPR variable in the model for 1977-78. Several other firms. Alternative explanatory factors were sought when specifications without WPR were analyzed for the plas- evidence was found that government incentives do not tics industry, however, and the results for four of them affect a firm's location. In particular, it was found that are presented in table 9-9. An analysis of the coefficients entrepreneurs attach an insignificant weight to govern- for the variable Fl shows that for plastics the financial ment financial incentives. incentive variable enters in all four specifications with a The model does not claim to represent completely the large and significant coefficient. Thus, for this sector complex set of locational factors that enter into the financial incentives exerted a substantial effect on loca- decision process; it only attempts to point out some tion. One might argue that these results are a product of factors that decisionmakers seem to consider when sampling error. On the assumption that they are not, we choosing a location. Some of these may be subject to offer the following explanation, based on interviews with control by policymakers. entrepreneurs. isic code 356 includes the molding, ex- The findings should be interpreted in the context of truding, and fabricating of plastic articles. A typical two limitations. First, only a short period has elapsed factory has few employees, requires relatively little floor since 1974, when the first presidential decree concern- space, and has one or two machines that process raw ing industrial deconcentration was issued. Second, the material. The basic component of the machines is the data base of this study cannot be accepted without the mold, which determines the end product (such as a fork, standard caveats, and the results must be considered plate, or syringe) and accounts for a high proportion of tentative. This study reaches cautious conclusions on Industrial Deconcentration Policy: Venezuela 119 the basis of available statistics and indicates a methodo- Cordiplan. 1973. Encuesta Industrial 1971. Caracas. logical approach for future work. . 1979. Evaluacidn de la Politica de Desconcentracidn Industrial. Caracas. Fair, Gordon, and John Geyer. 1954. Water and Wastewater Notes Disposal. New York: John Wiley & Sons. Ministerio del Ambiente. 1979a. Esquema de Ordenamiento 1. Data are author's estimates based on the matched direc- de la Region Centro Norte Costera. Vol. diagn6stico, tomo tories and Oficina Central de Estadistica e Informatica (1978). de sintesis. Caracas. 2. World Bank data; Oficina Central de Estadistica e Infor- . 1979b. Esquema de Ordenamiento de la Region Cen- matica (1978); matched directories for 1974 and 1976. tro Norte Costera. vol. 7: Industria. Caracas. Throughout this chapter, an industry group is classified as Murray, Michael P. 1982. "Here, There, Where? A Strategy for having a low level of protection if its effective protection is Evaluating Industrial Relocation Policies in Korea." Urban below 80 percent and as having a high level if effective protec- Development Discussion Paper 6. World Bank, Washington, tion is above 130 percent. D.C. Oficina Central de Estadistica e Informatica. 1974. Encuesta Bibliography Industrial. Caracas. 1976. Encuesta IndustriaL Caracas. Carlton, Dennis. 1979. 'Why Do New Firms Locate Where . 1978. Directorio Industrial. Caracas. They Do: An Economic Model." In William Wheaton, ed., Interregional Movements and Regional Growth, pp. 13-50. Washington, D.C.: Urban Institute. Part IV Managing the City Part IV is devoted to a detailed discussion of specific cially to growing numbers of poor people. Positive in- urban problems. We have traced the roots of urban- come elasticities for public services and demonstration ization concerns and reviewed the limits of aggregative effects from the developed world also influence the de- concentration and decentralization policies; now we mand for local services. Factors that affect the costs of turn to more sectoral policy issues. Bahl and Linn in local services include the increasing costs of loans for chapter 10 address the fiscal problems of city manage- large urban infrastructure investments, wages, labor ment. Problems of housing, transport, and the environ- unions, rising land and energy prices, and inflation. ment are covered by Ingram, Pachon, and Thomas in The analytical and empirical findings presented in chapters 11, 12, and 13, respectively. Evaluation of chapter 10 are relevant to most developing countries. urban projects is discussed by Keare in chapter 14. The Efficient means of raising taxes and controlling expendi- discussions directly address policy issues to reveal the ture are explored. In many urban contexts a greater use scope for ameliorating urban problems. of property and motor vehicle taxes, to be adopted care- An important aspect of urban finances is their im- fully by local authorities, could be effective. Greater plications for alleviation of urban poverty. In principle, a reliance on user charges has the merit of being directly progressive tax policy can favor the poor, although a linked to services provided. The chapter also reviews greater impact on poverty can probably be obtained from controversial proposals and discusses political consid- expenditure measures in such areas as housing, trans- erations. Reform proposals are commonplace; concrete port, and social services. There are, however, extensive actions have been rarer, for a variety of reasons men- examples of failed initiatives in these areas-slum re- tioned in the chapter. It could be inferred from develop- moval, the banning of street vendors, and expensive ing-country experience that gradual adjustments might investment programs. Urban investments, if they are to be more realistically expected, and perhaps more sus- contribute to the welfare of the poor, must take into tainable, than radical and sweeping reforms. account the needs of the poor and their ability to afford the services. Urban Housing Toward Greater Fiscal Efficiency Overcrowding and lack of adequate housing are widely Chapter 10 provides a comprehensive statement on observed urban phenomena. In response, many govern- urban revenue and expenditure for developing coun- ments attach high priority to the provision of urban tries. The growing fiscal problems in cities in developing shelter to low-income groups. Most analysts now recog- countries can be attributed to unprecedented urban nize that housing includes not only shelter structures growth, for which many countries are ill-prepared, and but also the land and the services provided, including to the demands made on local services by particular new water and sewerage. Urban policy toward the provision investments for economic development. The demand for of such public services has a great impact on the quality local services is sensitive to increasing population, espe- of life, comparable to the provision of shelter itself. 121 122 Part IV The importance of housing is borne out by the high costs contribute to the configuration and growth of proportion of household budgets that is spent on it. cities, and transport often turns out to be a critical World Bank estimates indicate that average housing constraint on further expansion and on the smooth expenditures in developing countries range from 15 to operation of the city. Government policy on urban trans- 25 percent of income. Furthermore, according to U.N. port has important effects on the quantity, quality, and data an average 4 percent of the GDP in developing coun- price of this service. Policy issues include government tries is spent on new residential construction. Of course, intervention regarding licensing, public versus private there is considerable variation in these estimates among service, fares and subsidies, routing decisions, and in- countries, and the quality of data varies widely. vestments. The differences between urban and rural areas bring Careful estimation of supply and demand for transport out the origins of the urban housing problem. Housing is necessary for carrying out investments to improve expenditures are systematically higher in urban centers service. Transport demand and costs vary, roughly, than in rural areas. Thomas (1978) shows that the aver- directly with city size, which implies that transport age household's nominal expenditures in Lima may be policy is increasingly important as urbanization pro- three times higher than in the rural areas of Peru. ceeds. The impact of poor transport is particularly seri- Similar findings are contained in a study on Brazil by the ous for the poor who commute to work. The study of same author (Thomas 1982). In general, much of the Colombia in chapter 12 draws attention to the need to variation in urban-rural expenditures seems to be ex- focus on cost-effectiveness and affordability in the for- plained by the disproportionately larger outlays in urban mulation of transport policy. Of particular interest are areas on nontradables, the most important of which is the implications regarding efficiency and equity in housing. several areas: public versus private transport, old versus The quality and quantity of housing in urban areas are new vehicles, large versus small buses, and govemment an important aspect of the urban problem. To improve interventions and subsidies that affect these choices. the housing supply, policymakers must take into account in greater detail the behavior of market demand and the cost of providing housing. The demand for housing reflects willingness to pay for a set of housing Urban Services and the Environment services. It is important to measure adequately the de- mand for housing in the context of the growth of cities, This volume does not elaborate on several of the income expansion, and changes in relative prices. important services that contribute to the quality of ur- Chapter 11 deals with one aspect of housing demand ban life: education, health care, nutrition programs, and provides an illustration of the equation for estima- family planning, and so on. Many difficulties in these tion of housing demand. The estimates are based on areas stem from the basic problem of urban poverty, and household interview data from Bogota and Cali, Colom- policies that stimulate developmentwould go a long way bia. A comparison of parameter values with those toward addressing them. Direct programs also have a obtained from North American data sets shows that role. It is important, however, that, as in the case of demand elasticities in Colombia are generally compara- housing and transport, emphasis is placed on observed ble in magnitude with those in the United States. The demand patterns, affordability, and the cost- approach employed to represent variation in housing effectiveness of policy instruments. price in the demand equations uses a theoretically A variety of amenities contribute to urban well-being. attractive and computationally straightforward proce- Chapter 13 deals with the impact of environmental dure that is based on residential location theory. A sim- pollution and the means of improving environmental ple exercise illustrates the magnitude of bias of the quality. Air, water, and noise pollution and solid wastes income elasticity of demand that can result from incor- are commonplace problems in the developing-country rect data aggregation techniques. Moreover, correctly metropolis. Damages from these sources affect health aggregated data produce income elasticity estimates and property. The chapter provides evidence of the ill that are similar to those obtained from disaggregated or effects of air pollution on human health, indicating one micro data. type of benefit from pollution abatement. The control of pollution is costly, however, and cost- efficient policy tools must be devised. Cost estimates for Urban Transport certain types of industries and locations are provided, illustrating criteria for differentiating among sources in Transport is both a factor that drives city growth and a antipollution policy. Generally, it pays to differentiate source of urban problems. Transport technology and between locations and pollution sources and to reduce Managing the City 123 pollution where it is relatively less costly to do so and Bibliography where the benefits of so doing are relatively large. Churchill, Anthony A. 1972. Road User Charges in Central America. Baltimore, Md.: Johns Hopkins University Press. Improving the Efficiency of Shelter Projects Dwyer, D. J. 1975. People and Housing in Third World Cities. London: Longman. Careful project evaluation has high payoffs in making Linn, Johannes F. 1983. Cities in the Developing World: Poli- government interventions, where they are justified, cies for Their Equitable and Efficient Growth. New York. efficient. Chapter 14 summarizes the experiences of four Oxford University Press. World Bank loan programs designed to improve shelter; Payne, Geoffrey K. 1977. Urban Housing in the Third World. the lessons apply to other areas as well. Successful shel- Boston: Routledge and Kegar Paul. ter improvement projects have striven for efficient re- Pfeffermann, Guy Pierre, and Richard C. Webb. 1977. The source use through decentralized decisionmaking. The Distribution oflncome in Brazil. World Bank Staff Working chapter evaluates project performance and stresses mar- Paper 356. Washington, D.C. ket solutions where they are feasible, on the presump- Thomas, Vinod. 1978. The Measurement of SpatialDifferences tion that project participants are the best judges of their in Poverty: The Case of Peru. World Bank Staff Working own self-interest. Paper 273. Washington, D.C. Also published in Journal of In general, projects should provide participants with Development Economics, vol. 7 (1980), pp. 85-98. suitable locations, secure tenure, and adequate credit . 1982. Differences in Income, Nutrition and Poverty but beyond these should leave most decisions to the within Brazil. World Bank Staff Working Paper 505. participants. The advantages and disadvantages of con- Washington, D.C. struction projects, self-help requirements, housing Walters, Alan. 1979. 3The Benefits of Minibuses." Journal of standards, rentals, and restricted credit policies are dis- Transpor) Economics and Policy, vol. 13, no. 3 (Septem- cussed. Also addressed are the costs of delayed occupan- cy and inadequate maintenance and the importance of project cost recovery. The findings underscore the importance of carrying out evaluations. 10 Intergovemmental Fiscal Relations in Developing Countries Roy Bahl and Johannes Linn The financing and management problems of many of creased productivity associated with it. It follows that if T the largest cities in developing countries have be- cities are unable to finance services, maintain their gun to attract the serious attention of economists. infrastructure, or accommodate population growth, na- Several factors explain why these issues have been so tional economic growth will be slowed. Second, there long ignored: the traditional concerns of development is increasing concern about the problems of the urban economists have been macro growth models and the poor and the need to provide them with improved living agricultural sector; the urban fiscal problem was small conditions. Water, sewerage, primary education, and in relation to the financial problems of the central gov- refuse collection are locally provided services to which ernment; aid donors dealt with central governments; the urban poor often do not have adequate access. Im- the fiscal data of local governments were scanty; and portant related problems concern issues of how to allo- troublesome local government issues seemed better left cate more resources to the provision of such services, to administrative specialists. Things have changed, pri- how to distribute a fair share to the urban poor, and marily because the fiscal problems of cities have become whether and how to charge recipients for their use of national concerns and because donors have recognized these services. that the success of capital projects in urban areas is Third, the resource mobilization issue has assumed closely tied to the ability of local governments to meet growing importance. In most developing countries the recurrent cost obligations. share of GNP that is mobilized for total public sector This chapter addresses an increasingly important activities is thought to be too low, and local govern- aspect of urbanization in developing countries-the ments may contribute measurably to increasing it. Ur- problems and practices of urban government finances. It banization generates increases in taxable capacity that identifies and analyzes the most important pressures on can sometimes be more easily reached by local than by local budgets, suggests major options for reforms, and central governments. Increases in property values can sets out constraints on improvements. An important be captured by property taxes and growing business limitation to this effort is the inadequacy of comparable activities by business taxes. Some expansions in formal data, as reflected by the paucity of empirical support for employment can be more easily tapped by local than by these arguments.' central income taxes, and automobile-related charges might be effectively levied by local governments. With respect to the provision of certain services, such as water Urban Government Finance supply, sewerage, and transport, there is a substantial as a National Problem opportunity for local governments to recapture costs through user charges. It would seem reasonable to That the financial health of cities is an important assume that increasing the mobilization of local govern- national issue for developing countries is borne out by at ment resources could have a significant effect on nation- least five considerations. First, economic growth at the al tax effort. national level is often led by urbanization and the in- Fourth, external lenders have begun to recognize the 124 Intergovemnmental Fiscal Relations in Developing Countries 125 importance of strengthening local government finances. is the quantity of service i required, P is population, ei is Capital projects have annual operating and maintenance EiliQ, the unit cost of the required service, and ti is costs, and in some cases the supporting services must be Qi IP, quantity of service i required for each inhabitant. provided. For example, a sites and services housing The actual level of expenditures for public service, Ei, project must have a water and sewer system, adequate may be defined as refuse collection, street lighting, road maintenance, and E Q access to primary education services. In many cases the (10-2) Ei= -'P= eiqiP provision of these services lies with the local govern- Qi P ment, and therefore the success of the entire project where Qi is the quantity of service i actually provided. depends on the ability of local government to meet the Local government revenues, R, may be defined as necessary recurrent costs. R=T+C+G Finally, questions have been raised about the optimal (10-3) distribution of city size. Some have argued that cities where T is taxes, C is user charges and other current have become too large, that the resulting fiscal and revenues, and G is externally raised revenues. management problems are insurmountable, and that The fiscal gap, D, in a city may then be defined as centralization of population and economic activity is (10-4) D = i - R = Y (ei4iP)-R somehow bad for economic development. Others have i E countered by noting that productivity advantages are By contrast, the actual budgetary deficit, D, is associated with city growth and that effective urban (10-5) D = Ei - R = Y (eiqiP) -R. management is possible with the correct institutional i E framework. Nevertheless, a decentralization movement This formulation clarifies the distinction between the is under way in many developing countries to increase fiscal gap, which reflects the shortage of revenue avail- the fiscal autonomy of local governments and to bring able to provide required services, and the budget deficit, government decisionmaking closer to the people. which reflects the actual shortfall of revenues. Budget deficits do not always occur, but fiscal gaps are common- place. Equations 10-1 through 10-5 are also useful in The Urban Fiscal Gap organizing the discussion of urban fiscal problems by focusing separately on the expenditure and revenue Urban governments in developing and developed sides of the local government budget. countries alike complain that their resources are in- adequate for providing sufficient urban services. The gap between perceived service needs and financial resources Expenditure Pressures (the "fiscal gap") has been interpreted in several ways. The most common explanation begins with the observa- The expenditure side of the local budget is subject to tion that urban populations in developing countries two sets of pressures: demand or needs which raise Qi, have expanded rapidly in recent years and are likely to and cost factors which raise ei. Much has been written continue expanding in the foreseeable future (World about how these factors affect expenditures in developed Bank 1979). Urbanization has led to rapid increases in countries (Bahl, Johnson, and Wasylenko 1980), but less expenditure requirements, but revenues have not in- attention has been paid to expenditure determinants in creased commensurately because local govemments are developing countries. One might ask whether the de- often restricted to income-inelastic revenue sources. cisionmaking model used to explain urban fiscal out- A better understanding of the nature of the problem comes in advanced countries fits the developing-country and the altemative reform possibilities requires some experience. understanding of the components and determinants of. The advanced-country model of expenditure deter- the fiscal gap. A useful starting point may be to cast the mination holds that the fiscal choices of politicians are problem in terms of a set of identities that define the influenced by the preferences of the median voter, the expenditure requirements and revenue constraints of relative prices of public goods and services, the income urban authorities. Expenditure needs or requirements level of the community, and the availability of external for the ith public service in a particular city may be resources. The developing-country case often differs be- defined as: cause voters have less chance to express their pre- Ei Qi A A ferences; local councils are as often appointed as elected, (10-1) I= - pP = ei4zP and the chief administrator of the city may be a central Qi P government employee with substantial autonomy. In where ,i is the required expenditure for service i, Qi addition, local government financial autonomy is quite 126 Roy Bahl and Johannes Linn restricted. For instance, it is common for the central ginal costs because of the greater depth required for govemment to place tight controls on local govemment tubewells or the greater distance to a catchment area. tax rate changes and borrowing practices and to impose Urban population growth rates in developing countries constraining mandates for service levels. Although local tend to lie considerably above national population fiscal choices in developing countries are more con- growth rates (table 10-1). Moreover, in some developing strained, however, they are by no means nonexistent. countries the rates of growth of the largest cities tend to Many local councils and mayors are elected; the com- be even higher than the growth of the total urban position of centrally appointed local councils often population. reflects local political considerations; and appointed city managers do attempt to take local preferences into Income Effects. The positive and strong relation be- account. Even in the most centrally planned and cen- tween urbanization and per capita income in developing trally controlled developing countries, the public pro- countries has been well established in two respects: the tests high bus fares or water rates, sometimes effectively resists tax rate increases, and often demands increased public services. Table 10-1. Urbanization in Selected Developing All of this implies that a more constrained version of Countries, 1960-80 the traditional maximization model would be relevant Urban population Percent of for developing countries. To this end it seems necessary as percent of urban population to explain the determinants of urban govemment spend- total population in largest city ing levels in terms of demand-related or need-related Country 1960 1980 1960 1980 factors, cost factors, and the capacity to finance. Low income' 15 17 11 13 India 18 22 7 6 Demand or Need Factors Sri Lanka 18 27 28 16 Pakistan 22 28 20 21 The demand for local public goods is determined by Tanzania 5 12 34 50 relative prices, incomes, needs, and taste factors. Tastes, Zaire 16 34 14 28 or preferences, affect the rate at which consumers are Indonesia 15 20 20 23 willing to substitute private for public goods. For exam- Sudan 10 25 30 31 ple, changing preferences may reflect the demand for Middle incomeb 37 50 28 29 better educational services by families whose income has Kenya 7 14 40 57 risen above subsistence levels, changing societal values Ghana 23 36 25 35 such as substitution of welfare and housing services for Egypt 38 45 38 39 the extended family structure, the demand for more Zambia 23 38 n.a. 35 redistributive actions to prevent unrest,2 and willingness Thailand 13 14 65 69 to pay more taxes in return for govemmental action to Philippines 30 36 27 30 offset negative extemalities that result from the growing PNera 436 67 38 397 underprovision of urban public services. Colombia 48 70 17 26 Cote d'lvoire 19 38 27 34 Needs. An important consequence of urbanization is Tunisia 36 52 40 30 that public service requirements change and decision- Jamaica 34 50 77 66 makers may have to interfere with or override individual Malaysia 25 29 19 27 preferences in providing them. The continuing increase Korea, Rep. of 28 55 35 41 in the numbers of the urban poor calls for increased Algeria 30 44 27 12 social and economic services and perhaps for a different Mexico 51 67 28 32 package of public services-for example, serviced sites Chile 68 80 38 44 rather than permanent housing, small health clinics Brazil 46 65 14 16 Iran 34 50 26 28 rather than hospital additions, more standposts rather Argentina 74 82 46 45 than water main extensions, and the like. Venezuela 67 83 26 26 The growth in the need for public services is most often associated with increasing population. Some Note: Countries are listed in ascending order of per capita GNP. would argue that expenditures must increase at least in a. Weighted average for all low-income countries given in World proportion to population to maintain even a constant Bank (1981). b. Weighted average for all middle-income countries given in World per capita level of service (equation 10-1). For example, Bank (1981). water system expansions may involve increasing mar- Source: World Bank (1981). Intergovemmental Fiscal Relations in Developing Countries 127 more urbanized developing countries tend to have high- high quality standards for water supply, water-borne er per capita incomes (Beier and others 1975; Renaud sewerage technologies to replace such traditional dis- 1981; Smith 1974), and per capita income in the largest posal techniques as night-soil collection, and construc- cities tends to be the highest in the country (Linn 1982). tion of incinerators or composting plants for solid waste There is less objective evidence on the relation between disposal to replace conventional recycling techniques. the increase in urban population and the increase in per All these can result in significant increases in expendi- capita income in urban areas. ture requirements (Linn 1982). Increasing per capita income tends to increase the per capita demand for services (q); the magnitude of the Migration and Poverty. Increased local government increase is dependent on the income elasticity of de- expenditures also result from the locational decisions of mand for locally provided services. Positive income elas- poor migrants, who typically swell the populations of ticities for urban services have been observed for water cities in developing countries. Often these migrants supply, electricity, telephone service, and solid waste squat on or purchase at cheap prices land that is diffi- disposal services (Linn 1982). These higher levels of cult to service owing to topography (mountainsides, consumption of utilities may be largely attributable to swamps, flood areas, and so on). As these settlements increased ownership of appliances that use water, elec- become relatively well established, the need to service tricity, or both (washing machines, radios, televisions, them increasingly becomes a political and humanitarian and the like), while the increased need for solid waste necessity for urban governments, and the result is sub- disposal may be associated with generally higher con- stantial expenditure requirements. sumption levels and reduced recycling in the home. The demand for motor vehicles is highly income- elastic, which implies that the demand for urban high- Cost Factors way infrastructure is positively related to per capita Local government expenditures may also rise because income.3 The demand for schooling is strongly corre- the cost of providing a given quantity of public services lated with household income, since the lower the in- rises-that is, because of an increase in ei. Increasing come the more likely it is that children will be forced to factor costs, particularly for labor, are a prime reason for drop out of school to seek employment (Beier and others increases in local government expenditures. Conversely, 1975), and the less able are households to bear out-of- local government unit costs may fall as population in- pocket expenditures for education (Meerman 1979). creases if there are economies of scale in the provision of Similarly, the demand for health care increases with urban public services. incomes, with better education, and with rising familiar- ity with modern health care techniques. Unit Cost of Inputs. Probably the principal factor Per capita income increases also result in a demand responsible for unit cost increases is inflation. To the for a higher quality of urban services. Higher quality extent that developing countries are plagued with higher may mean individual rather than communal water sup- levels of price inflation than are industrial countries, ply and sanitary facilities,4 reduced risk of electric out- local government costs in developing countries will in- ages, more rapid communication and transport, and crease more rapidly than those in industrial countries. better health care, education, and fire and police protec- General inflation can lead to increasing per capita ex- tion. penditures but does not explain an increasing share of local government expenditure in income. There are, Expectations and Demonstration Effects. Changing however, factors that tend to raise urban public service expectations regarding appropriate service levels and input costs more rapidly than the general inflation rate. quality also increase the demand for urban services. This First, the provision of public utility services frequent- is especially true for water- and electricity-using ap- ly requires investment outlays on a large scale. Urban pliances, motor vehicles, and educational achievement governments therefore have to rely on foreign credits and therefore for the derived demand for such related from international agencies, from the international urban services as public utilities, road construction, and capital market, or from suppliers. As is well-known, the education. The demand for public services may also be supply curve for funds is upward sloping, and large cities heavily influenced by a demonstration effect from more are likely to run into increasing costs of capital unless developed countries. As a result, developing-country the central government is the primary borrower and governments have frequently raised standards rapidly to passes on the loan funds at subsidized rates to urban attempt to attain the levels of quality and technology governments. The extent and terms of borrowing by found elsewhere. Examples are the often unchecked urban governments are almost always controlled by the growth of private automobile ownership, high use of and central government, and thus the cost of capital as it is 128 Roy Bahl and Johannes Linn passed on to local authorities is effectively a policy in- larger cities than in rural areas, although these differ- strument of the central government. The rising cost of ences are rarely measured accurately. Thomas (1978) capital is reflected ultimately in the public sector as a found that in Peru the average cost of living in Lima financial cost, whether at the national or at the local exceeded that in rural areas by a substantial margin. level, and thus the large capital-intensive infrastructure This may be largely explained by differentially higher works required by rapid urbanization may well involve input prices in Lima, compared with the rest of the higher-than-average interest rates. country. A similar conclusion for Brazil is presented by Second, the average cost of public employees may also Thomas (1982). be affected by urbanization, although the direction of this effect is not completely clear. If workers become Technological Progress and Economies of Scale. more productive as city size increases, and if private Technological progress should reduce input re- sector wages respond to productivity increases, there is quirements per unit of public output and thereby reduce some possibility for a wage rollout to the public sector. unit cost (ei). It is generally recognized, however, that This suggests increasing public sector wages during the technological progress in most public services is slow urbanization process. Yet the extent to which the com- (Baumol 1967) and that some technological advances petitive wage thesis holds depends on the rate of in- may lead developing countries to adopt modern tech- migration and, other things being equal, a greater rate nologies which are inefficient and excessively costly. of in-migration will dampen the wage rollout effect. The Examples are composter facilities for solid waste dispos- local public sector, as part of the general services sector, al, waterborne sewerage systems and treatment plants, is labor-intensive and does not require a skilled work limited-access rapid speed highways, subways, and pos- force. Newly arrived migrants swell the numbers of sibly even premature computerization. unskilled workers available and hold down the wage rate Technological economies of scale may also imply de- in the services sector. clining unit cost during urbanization. A detailed study of It may also be argued that wage rates increase more engineering costs for certain urban services in India rapidly than does the general price level, owing to insti- appears to support this hypothesis (Stanford Research tutional factors that determine the wages and salaries of Institute 1968). Similarly, a study of water supply costs local government employees. Labor unions for local civil for small and intermediate-size cities in Colombia has servants exist in some countries and can be extremely shown declining unit costs (Insfopal 1975), which are vocal in pressing for higher wages.5 In other cases local probably largely attributable to economies of scale. government salaries and wages are determined by the These economies of scale, however, are likely to be central govemment. Since it frequently does not bear limited to public utilities, and it is not even clear the brunt of local government salary readjustments, the whether governments in rapidly growing urban areas central government may be quite willing to raise local are always in a position to benefit from them. Moreover, civil service salaries more rapidly than the general price economies of scale may be offset by diseconomies of level.6 agglomeration, particularly in larger, denser urban Third, land prices are also likely to rise more rapidly areas. Diseconomies result from congestion, which than the general price level in cities as a result of rapid tends to increase with city size and is especially prob- urban population growth, increased density, and the lematic in transport. Diseconomies may also result from resulting scarcity of serviced urban land. Rising land limits in the carrying capacity of the natural environ- prices tend to have particularly strong impacts on the ment (examples are air and water pollution) and from unit costs of services that are relatively land-intensive, the increasing scarcity of natural resources, especially such as the transport sector-which requires substan- water and energy. Other examples are the increased tial amounts of urban land for streets and sidewalks- need for disease control and fire and police protection parks and recreation, schools, and solid waste disposal. which are associated with the large scale and high densi- Finally, the unit cost of energy has risen more rapidly ty of urban living. In all cases some inputs (space, natu- than the general price level and may well continue to do ral resources, and so on) grow scarce as urbanization so. This can burden larger local governments that tend proceeds, and larger amounts of other inputs (labor, to rely more heavily on the use of motor vehicles for capital, and intermediate inputs, in particular) have to provision of services and on the use of electricity-for be applied to maintain service levels and provide clean instance for street lighting and for the pumping and water and air, good health care, and a safe environment. treatment of potable water and of sewage. Confirmation of the hypothesis that input prices tend Public Service Employment. Cost increases for to increase with the population size of cities is found in urban services may result if local government is viewed frequent observations of price levels that are higher in as an employer of last resort. Overstaffing of local gov- Intergovernmental Fiscal Relations in Developing Countries 129 ernment offices is typical, and as a result the financial rate ceiling is usually higher for larger cities. Thus, costs of urban services are inflated. Whether this is urbanization may give a substantial boost to local gov- desirable as a matter of policy depends on the tradeoff ernment tax revenues through increases in the values of between the gains from additional employment, often at Y, P, rj, and tj. low economic costs (because of widespread unemploy- This favorable impact on local government revenues ment or underemployment of skilled and semiskilled can, however, be offset by important constraints. Local labor), and the financial impact of such a policy. governments may have little opportunity to capture the increases in taxable capacity associated with urbaniza- tion. For example, the central government fixes Bj by specifying the tax bases available to local governments; Revenue Constraints and Opportunities t1, the tax rate, is commonly fixed by the central gov- ernment; and in some cases even tax collection, rj, is Urbanization most often puts pressure on local gov- largely a central government matter. In sum, there is no ernment budgets by driving up expenditures; at the correspondence between the expected increases in ex- same time revenues may rise by less than a commensu- penditures and in revenues in response to a given in- rate amount. Following equation 10-4, total revenues crease in urbanization. The resulting gap or deficit will available to finance the growing expenditure require- vary from place to place and will depend on the public ments of cities may be separated into tax revenues (T), service responsibilities and discretionary fiscal powers of other current revenues, including user charges (C), and local governments. external funding (G). User Charges tares The revenues generated by user charges (C) may be Total tax revenues of an urban government are deter- represented as mined by a set of factors which may be summarized in C Q the definitional identity (10-7) C ___ - t (c q1P) (10-6) T = E ( Ti j pj P)= rj tj b1 yP where Ci is user charges collected for urban service i, Qi -L1B1YP is the quantity of service i consumed, ci is the unit where charge for service i, and qi is the quantity of service i T = total tax revenue provided per capita. Tj = tax revenue of taxj In contrast to urban taxes, user charges for services Lj = legal tax liability of tax] for given tax statutes show a direct link between the quantity of services pro- Bj = base of tax] vided and the revenues generated to finance those ser- Y = total personal income vices. The extent to which user charges cover the cost of rj = collection rate provision of services depends, however, on how the tj = legal tax rate average price charged, ci, compares with the average bj = base to income ratio cost of providing service, e,. y = per capita income The evidence on the relations between city size and P = population the cilei ratio is mixed. On the one hand, autonomous public utility agencies in large cities seem able to charge Considering the rapid growth in population and in- high enough rates to cover increasing marginal costs come and the relatively high per capita income levels and sometimes to generate a surplus. The same result associated with urban growth, the process of urbaniza- seems to hold for special assessments on urban land- tion should and does provide a relatively strong taxable owners, for example, in Korea (Doebele 1979) and Co- capacity for urban governments. Not only do increased lombia (Doebele, Grimes, and Linn 1979). Urbanization population, income, and motorization represent a thus seems to create a demand for these services and a strong economic base (in relation to the rest of the capacity to pay full costs. On the other hand, some country), but larger cities also have the potential for services (notably transport and housing) do not generate more efficient collection. Urban local governments have enough revenue from charges to cover their full costs. a better professional expertise than other local govern- The problem here is that urbanization also generates a ments, and urbanization brings growing formal employ- great many social costs, such as congestion and pollu- ment, more cars, and increased property values-all tion, and poverty problems that may dictate holding c1 readily identifiable tax handles. Finally, the statutory tax below ei. 130 Roy Bahl and Johannes Linn External Funds certain urban services such as public utilities, roads, education, and health. In addition to arguments of re- Grants and loans (G) are not under the control of local lieving financial pressures, efficiency and distributive authorities but depend on the decisions of higher-level concerns are often cited as justifications for such reas- authorities. Whether urbanization brings more grant signment. Because of the difficulties of measuring the assistance to local governments depends on how the efficiency and equity gains, however, it is difficult to revenue sharing system operates. For example, if grants justify such proposals. In any case, two important con- are distributed according to population or on the basis of siderations weigh heavily against this approach to re- local tax collections, as in a shared tax, urbanization may solving urban fiscal problems. First, the transfer of spe- generate an increased inflow of external resources. The cific functions to higher-level government reduces the same may be true if grants are made on a cost reimburse- local authority's potential for responding to urban policy ment basis. issues and problems. This is undesirable because many In the functional relation G = G(P, Y, Qi), one would urban development activities are interrelated and re- usually expect to find partial derivatives such that quire an integrated approach to planning and im- dGIdP 2 0, dGIdY $ 0, and dGldQi ˇ 0. plementation, and local authorities are often better In other words, grants tend to vary directly with city equipped to provide such planning than are national population size and with the amount of services pro- ministries or special-purpose agencies. Second, national vided under a system of per capita or cost reimburse- governments often assume only the responsibility for ment grants. External resource flows may increase or making capital investments and leave it to the local decrease in response to increases in per capita income in authorities to operate and maintain the facilities. This the city, depending on the grant system. A tax sharing turnkey approach tends to burden local authorities with scheme will channel more funds to cities as urbaniza- facilities that are often beyond their financial and tech- tion proceeds, whereas a formula that equalizes grants nical capacity to operate and maintain and that may not across jurisdictions may have the opposite effect. reflect local preferences. Policy Options for Urban Fiscal Reform Increased Local Taxes The judgment about which taxes are most appro- It may be concluded with some certainty that public priately allocated to local authorities depends in part on expenditure requirements increase with urbanization in the perspective of the decisionmaker. From the central absolute terms and probably also in per capita terms. government's perspective the main goals are to (1) limit Most often, urbanization also enhances the revenue local competition for the important national tax bases capacity of urban governments, but revenue growth in (broad-based wealth, income, and expenditure taxes); most cities has been hampered by the limited revenue (2) limit the local use of taxes that are mainly exported to authority granted to local governments and by their other jurisdictions; (3) provide local authorities with a poor revenue efforts. As a result, revenues have not kept reasonably buoyant revenue base; (4) avoid local reliance pace with expenditure needs, and service deficits have on regressive taxes; (5) encourage the use of taxes that resulted. There is little reason to expect that this situa- are most easily administered at the local level; and (6) tion will change significantly in the future. encourage the use of taxes which are closely linked to This conclusion raises the question of the constraints urban infrastructure and congestion costs, to internal- on reform. In principle, an urban fiscal gap can be closed ize some of the externalities in the urban economy. in four ways: increased local revenue effort with un- From the local perspective, criteria 3, 4, 5, and 6 are changed revenue authority, increased local revenue au- likely to be equally relevant, although they may vary in thority, increased transfers from higher-level govern- strength. For example, local governments are likely to ment, or reduced responsibility for expenditures. With place a greater emphasis than higher-level governments these alternatives in mind, the reform question may be on buoyancy and administrative ease and perhaps a addressed by focusing separately on the expenditure and different emphasis on equity and efficiency. As regards revenue sides of the budget and by noting important criteria 1 and 2-competition with national tax bases interrelations between the two sides. and nonexport of revenues-local authorities are likely to have priorities exactly opposite to those of higher- The Reassignment of Functions level government. Since the broad-based taxes tend to be the more buoyant, and the most easily tapped, local A common response of higher-level governments to governments desire access to them. Reliance on taxes city fiscal crises has been to assume responsibility for which can be shifted to taxpayers outside the jurisdic- Intergovernmental Fiscal Relations mi Developing Countries 131 tion will naturally also be politically attractive to local emphatic in their opposition to increases and at times govemments. even endanger the political stability of the country Given these sometimes contradictory goals, it appears through riots and the like. Local authorities also have that the property and motor vehicle taxes are, on bal- mixed attitudes about raising user fees, partly for the ance, the most desirable and least objectionable of the same political reasons that concem the national govern- major tax instruments that could be delegated to local ment and partly because they may have come to rely on jurisdictions. From the central perspective, they do not central govemment transfers to finance significant por- compete substantially with national taxes, and tax ex- tions of their public service investments. porting is likely to be limited, particularly for large On balance, however, it appears that local service cities.7 From the local perspective, too, these taxes are charges have become increasingly important sources of largely appropriate. Local access to broad-based con- additional revenue, particularly in cities in which local sumption, income, and wealth taxes is generally not authorities provide public utility services. In many cases granted by national governments-a reflection of the user charges have been utilized effectively by local au- overriding influence of central government objectives. thorities in lieu of alternative revenue sources and have Exportable taxes, such as selective excise taxes, octroi (a contributed to the general financing of the local govern- tax on goods brought into a town), and tourism and ment. It is in the area of user charges that judicious hotel taxes, are popular among local authorities but tend support by central authorities for local revenue raising to receive only mixed blessings from the central author- efforts, in place of the frequently practiced obstruction- ities or to be prohibited. They are sometimes tolerated ism, may be most productive in closing the urban fiscal mainly because they reduce local govemments' claims gap. on national tax resources. The empirical evidence on the actual use of taxes by , local governments in the cities of developing countries Grants is fully compatible with the gist of this discussion (Bahl, An increase in fiscal transfers from the central to local Holland, and Linn 1982). The practical difficulty is that govemments is another means of coming to grips with the scope for transferring additional taxing authority to the urban fiscal gap. There are a number of good reasons local governments may be severely circumscribed by for developing a system of revenue sharing between national economic development policies. Rather than different levels of govemment: the greater administra- expect the allocation of new revenue authority, it may be tive efficiency of collecting taxes centrally, the need for more realistic to argue for reduced central government the central govemment to work to equalize the revenues limitations on the use of taxes already collected by local of local governments, and-the traditional justifica- govemments. Such policies, together with local mea- tion-the need to provide incentives for efficient local sures for the improvement of local tax effort, will permit govemment decisions on resource allocation. For large local government to better reach the growth in taxable cities, however, it is unrealistic and inappropriate to capacity that comes with increased urbanization. expect that transfers will reliably and permanently fill the fiscal gap. The pressures on the central fisc tend to be User Charges such that transfers to local authorities are the first to be cut when national austerity programs are implemented. From the national perspective, an effectively adminis- tered set of user charges would seem to be the most appropriate source of local government revenue. User The Politics and Prospects charges do not compete with central government rev- of Urban Fiscal Reform enue bases; they are largely nonexportable; they can have desirable revenue, efficiency, and equity charac- Reform proposals to alleviate the fiscal problems of teristics; and they are administratively feasible at the urban govemments are commonplace. Although the local level. It may therefore come as a surprise that nature of these proposals has varied from place to place, national govemments have sometimes counteracted in line with local conditions and with the makeup of the local authorities' intentions to mobilize more resources study team, it is clear from a review of the evidence on through increases in user charges. There are two urban public finance reform that the proposals have far reasons for these interventions: national governments outnumbered the reforms.8 Major local government are concemed about inflation and want to limit the fiscal reforms in the last twenty years have taken place impact of rising local public service charges, and they mainly in the industrial countries. Examples are Ger- fear the political repercussions of price increases for many (consolidation of communes and reform of rev- urban services, since urban consumers are often quite enue sharing arrangements), Sweden (consolidation of 132 Roy Bahl and Johannes Linn communes), and the United States (reform of revenue ministry of finance, the ministry of local government, sharing arrangements). Among the developing coun- and the city governments. tries the rule would seem to be that changes in urban The ministry of finance, which usually is the strongest finance arrangements come slowly and that it may take party, tends to argue in favor of the status quo. It gener- decades for fundamental changes to take place, if they ally refuses to relinquish control over major tax sources occur at all. or borrowing and argues that decentralization would Typical for the developing countries are minor and compromise the central government's important fiscal slow adjustments in urban finance practices, such as the and tax policy programs. On the expenditure side, the creation of special districts for capital cities with special ministry of finance would rather emphasize central gov- expenditure responsibilities and revenue authority (for emient projects and priorities and is often suspicious example, Manila and Seoul), enlargement of metropoli- of the ability of the ministry of local government to tan jurisdictions by annexation of adjacent municipali- regulate the fiscal operations of local governments. It ties (as in Bogota), gradual development of new revenue will agree to a grant system but would prefer the grant sources (for example, betterment levies in Colombian pool to be decided on a year-by-year basis rather than cities, land readjustment schemes in Korea, and vehicle take the form of a shared tax. In general, the finance taxation in Jakarta), gradual reform of existing revenue ministry looks on local governments as junior partners sources (property taxation in Jakarta and Manila), reas- in the fiscal process, is doubtful about their manage- signments of expenditure functions (Kenya and Zam- ment and tax collection abilities, and feels that they can bia), and similar gradual and ad hoc responses to urban get much more out of their existing revenue authorities fiscal pressures. Major reform proposals have often been without recourse to new sources of revenue. shelved or taken up only in very minor respects-exam- The ministry of local government-in some countries ples are local government reform proposals in India and the ministry of the interior-is usually less influential property tax reform in Jakarta. Where major readjust- than the finance ministry and is often less well staffed. It ments in the fiscal structure of urban governments have usually argues for an extensive grant system and for occurred, it was where higher-level governments took other regulatory mechanisms which allow a greater over important revenue sources previously allocated to measure of control over local government finances. It local authorities (Kenya and Iran), where sweeping po- would prefer that the total grant pool be determined as a litical changes resulted in major shifts of national policy fixed share of some national tax and that the distribution priorities (Nigeria, Tanzania, and Uganda), or where the of some or all of these grants be at its discretion. Such a fiscal problems became so unmanageable that some scheme would at the same time limit the susceptibility drastic reform was unavoidable (removal of most impor- of the grant system to changing finance ministry budget tant expenditure responsibilities from rural and small- priorities and maximize the extent of the control of the town councils in Kenya). The lessons from the history of ministry of local government over local authorities. urban fiscal reform proposals and implementation are Administrations of large cities would prefer more in- that major proposals rarely have a chance of adoption dependent taxing power and less central regulation of and implementation and that gradual, stepwise adjust- their finances. If there is a grant system, a shared tax ments of the existing structure toward a more desirable based on origin of collection would be the preferred form state are all that can be hoped for. Gradual adjustments and a grant pool determined by the finance ministry and may, in fact, have a better chance of eventual imple- allocated by the ministry of local government would be mentation. less preferable. There are three major reasons for the inertia typically The sense of competition, the suspicion, and the lack found in the face of the need for urban public finance of mutual confidence that frequently characterize the reform. First, policymakers and citizens share an antip- debate between national and local government author- athy to the uncertain effects of untested large-scale ities in developing countries have constantly compro- changes in the economic environment. Second, most mised the potential for success of virtually any far- major reforms are associated with substantial windfall reaching urban government reform. If reforms are to losses to relatively few among the urban population- succeed, a mutually supportive system of local and cen- mostly among the elites-whereas windfall gains are tral government relations must be established. likely to be spread over a much larger number of peo- ple-mostly the less well-off. Third, although there has been a growing concern in developing countries about Notes how to strengthen the capability of urban govemments to come to grips with their tasks, progress has tended to 1. Perhaps the best comparable fiscal data available for become bogged down in a three-way debate over the developing countries are those reported in International Mon- fiscal decentralization issue that typically involves the etary Fund, Government Financial Statistics. Even this year- Intergovernmental Fiscal Relations in Developing Countries 133 book is more complete for the central government than for Baumol, William. 1967. "The Macroeconomics of Unbalanced local governments; in many cases, local government activity is Growth: The Anatomy of Urban Crisis."American Economic not reported at all. Revieu, vol. 57 (June), pp. 415-26. 2. It might be argued that utility functions become more Beier, George, Anthony Churchill, Michael Cohen, and Ber- strongly interdependent during the development process, at trand Renaud. 1975. The Task Ahead for the Cities of De- least partly because there is more to protect from the dangers veloping Countries. World Bank Staff Working Paper 209. of civil unrest. For the basic model see Hochman and Rodgers Washington, D.C. Also in World Development, vol. 4, no. 5 (1969). (May 1976), pp. 363-409. 3. See Smith (1974) for evidence that the density of auto- Doebele, William. 1979. "Land Readjustment as an Alternative mobile ownership is much higher in the major cities of de- to Taxation for the Recovery of Betterment: The Case of veloping countries than in each country as a whole. Fur- South Korea." In Roy Bahl, ed., The Taxation of Urban thermore, Smith and Kim (1979) show that the density of hopertyinLessDevelopedCountries, pp. 163-90. Madison: automobile ownership is higher in Seoul and Pusan than in the University of Wisconsin Press. intermediate-size cities of Korea. Data in World Bank (1975) also indicate a positive association between automobile own- Doebele, William, Orville Grimes, and Johannes Linn. 1979. ership and per capita income for a cross-section of cities in "Participation of Beneficiaries in Financing Urban Services: developing countries (see Linn 1983, ch. 4). Valorization Charges in Botoga, Colombia." Land Eco- 4. Note, for instance, that in the higher-income Latin nomics, vol. 55, no. 1 (February), pp. 73-92. American countries communal facilities are generally rejected Hochman, H. M., and J. D. Rodgers. 1969. "Pareto Optimal by the population, whereas they have found acceptance in Redistribution." American Economic Review, vol. 49 lower-income Asian and African countries. (September), pp. 542-57. 5. In Colombia, for instance, local teachers and health ser- Insfopal. 1975. "Metodologia para la formulaci6n del Plan vice personnel were extremely active in the 1970s in pushing Nacional deAcueductosyAlcantarillados." Bogota: Instituto for higher wages; they resorted to strikes, protest marches, and de Fomento Municipal. sit-ins at the municipal offices. Linn, Johannes F. 1980. "Property Taxation in Bogota, Col- 6. A good example is Korea, where the central government ombia: An Analysis of Poor Revenue Performance." Public raised local government wages and salaries by 20 percent in Finance Quarterly, vol. 8, no. 4 (October), pp. 457-76. 1975 and by a further 30 percent in 1976, both times signifi- . 1982. "The Costs of Urbanization in Developing Coun- cantly above the general inflation level (Smith and Kim, 1979). tries." Economic Development and Cultural Change, vol. Other examples of centrally decreed salary increases for local 30, no. 3 (April), pp. 625-48. officials have occurred, for example, in Kenya, Turkey, India, - 1983. Cities in the Developing World: Policies for and Pakistan. In Pakistan, however, the central government Their Equitable and Efficient Growth. New York: Oxford offset the increased expenditure requirements by providing a University Press. special grant to local authorities. Also, there are documented Meermany Jacob. examples ~ ~ ~ in whc.oa aaishveicesdlsail Meerman, Jacob. 1979. Public Expenditure in Malaysia: Who examples in which local salaries have increased less rapidly BeeisadM?NwYr:Ofr nvriyPes than the general price level, as in the case of municipal teachers in Bogota, where salaries, adjusted for increases in the Renaud, Bertrand M. 1981. National Urbanization Policy in price level, declined between 1971 and 1973. Developing Countries. New York: Oxford University Press. 7. Note, however, that opposition from groups that are Robson, William A., and D. E. Regan. 1972. Great Cities of the influential in national government circles has been known to World: Their Government, Politics and Planning. 3rd ed. hamper the development of effective property and motor vehi- London: George Allen & Unwin. cle taxation because of its progressive impact, which tends to Smith, Roger S. 1974. "Financing Cities in Developing Coun- be bome by precisely these groups. tries." IMF Staff Papers, vol. 21, pp. 329-88. Washington, 8. Walsh (1969); Robson and Regan (1972). D.C.: International Monetary Fund. - Smith, Roger S., and C.-I. Kim. 1979. "Local Finances in Non-Metropolitan Cities of Korea." Urban and Regional Re- port 79-1. World Bank, Washington, D.C. Processed. Bibliography Stanford Research Institute. 1968. "Cost of Urban Infrastruc- ture for Industry as Related to City Size in Developing Bahl, Roy W., Daniel Holland, and Johannes Linn. 1982. "Ur- Countries: India Case Study." Stanford, Calif. Processed. ban Growth and Local Taxes in Less Developed Countries." Thomas, Vinod. 1978. TheMeasurement of SpatialDifferences In Occasional Paper 57, Metropolitan Studies Program, in Poverty: The Case of Peru. World Bank Staff Working Maxwell School, Syracuse University. Syracuse. N.Y. Paper 273. Washington, D.C. Also published in Journal of Bahl, Roy W., Marvin Johnson, and Michael Wasylenko. 1980. Development Economics, vol. 7 (1980), pp. 85-98. "State and Local Govemment Expenditure Determinants: . 1982. Differences in Income, Nutrition, and Poverty The Traditional View and a New Approach." In Roy Bahl, within Brazil. World Bank Staff Working Paper 505. Jesse Burkhead, and Bemard Jump, Jr., eds., Public Em- Washington, D.C. ploymentandStateandLocal GovernmentFinances. Cam- United Nations. 1980. "Human Settlements Finance and Man- bridge, Mass.: Ballinger. agement." Theme paper for the third session of the United 134 Roy Bahl and Johannes Linn Nations Commission on Human Settlements, Mexico City, . 1979. World Development Report 1979. New York: May 6-14. New York. Processed. Oxford University Press. Walsh, A. H. 1969. The Urban Challenge to Government. New . 1981. World Development Report 1981. New York: York: Praeger. Oxford University Press. World Bank. 1975. Urban Transport. Sector Policy Paper. Washington, D.C. 11 Housing Demand in the Developing-Country Metropolis Gregory K. Ingram A policymaker who realistically hopes to cope with Developing countries can take several steps to more urban housing problems should understand hous- effectively meet the housing needs of their cities. Most ing demand and supply conditions in the city. Effective important is the identification of those tasks best accom- policy requires knowledge of how markets allocate re- plished by municipal governments and those best left to sources to housing, how homeowners and renters bid private developers and service suppliers in the housing for dwellings, how developers and contractors respond market. to housing demand, and how government regulations Some services are most efficiently provided by the and actions stimulate or constrain housing market municipalities. Electricity, water, and sewerage utilities activity. Our knowledge about many of these phe- have substantial scale economies that public corpora- nomena is modest, but our current understanding of tions may be best suited to provide. In addition, public housing markets in developing-country cities supports health can be effectively protected only if everyone par- several policy generalizations. ticipates in proper waste disposal and disease prevention The urban poor in developing and developed countries programs. Government services provided to all in such alike are the most likely to suffer from misallocated cases promote the public good. Local authorities, by housing resources. Slum removal worsens housing con- establishing and enforcing laws, can also define property ditions for the poor unless affordable low-cost dwellings rights and facilitate the transfer of property. In develop- are provided as substitutes for destroyed units. Often, ing countries customs regarding definition of property new units must satisfy quality standards for space, struc- rights may hamper commercial and residential land use. ture, and services, which substantially add to construc- In some developing countries unequal treatment under tion costs. Unless cities underwrite large housing sub- the law may bar groups from participating in urban sidies-which are beyond the fiscal means of most city development. budgets in developing countries-most new construc- Many well-intentioned municipal policies should be tion that meets high quality standards will be affordable avoided because of their perverse effects. Building codes only to middle- and high-income families. City mana- inform tenants of housing quality, but they can be harm- gers must recognize that housing goals that are set too ful if requirements are set too high. Zoning effectively high translate into no housing for those who have the bars developers from providing low-income housing in lowest incomes. many sections of cities. Rent controls discourage prospective developers by reducing future return on housing investment, and existing structures fall into Note: The author has benefited from comments and discussions disrepair if the returns from controlled rents do not with Michael Hartley, Stephen Mayo, Janet Pack, Joseph de Salvo, allow for maintenance expenditures. Similarly, usury Peter Schmidt, and participants in seminars at Princeton University, Michigan State, MIT, the World Bank, and the Corporaci6n Centro laws that prevent banks from charging rates that cover Regional de Poblaci6n. He thanks Sungyong Kang for research assis- their costs for small loans can effectively bar poor people tance, Maria Elena Edwards for manuscript preparation, and the staff of the Departamento Administrativo Nacional Estadfstica, Colombia, from formal housmg finance and leave them with the for aid with the data. options of progressive construction or informal finance 135 136 Gregory K. Ingram for housing. Municipalities can, however, sponsor credit quality, location, public services, and neighborhood programs when scale economies are involved and when amenities that are obtained in a single, tied purchase. private financial intermediation is primitive. Risk shar- Since there is no widespread agreement as to how to ing can extend credit opportunities to those for whom measure the quantity of housing, one analyst's price such options previously did not exist. variation may be another analyst's quantity variation. Private developers should be encouraged to provide Finally, even if we can agree that housing prices may for the housing needs of all urban dwellers. With higher vary, it is not obvious that all price variation should be incomes from urban employment and with expanded included in a housing demand equation. For example, if credit opportunities, the urban poor can finance their a metropolitan area's housing prices vary with the loca- own housing when standards are set appropriately. tion of housing but households can locate anywhere, we Some cities may decide to subsidize urban housing, cannot simply put the price actually paid by the house- which would enhance the resources at the command of hold into the demand equation, because the household the poor. Fundamentally, though, sound urban housing faces the whole schedule of prices. Simple inclusion of policy should be rooted in practices that are self- price indexes in a demand equation requires that house- supporting if housing is not to be a drag on urban holds be in different market segments. development. Numerous approaches have been employed to deal The scope for analysis of housing issues is obviously with one or more of these difficulties. Some examples broad, and this chapter addresses only one of many follow. issues. It looks exclusively at urban housing demand and 1. Assume that intrametropolitan price variation provides quantitative estimates of the factors that appear does not exist, so that all variations in expenditures to determine housing demand. Such quantitative esti- reflect variations in quantities; use expenditures in de- mates are required to design policies that permit the mand analysis as an index number to measure quantities supply of housing which is affordable for the urban poor. (Muth 1969). Three sets of results related to the estimation of hous- 2 Allow intrametropolitan prices to vary across ing demand equations are reported. First, estimates of . - housing demand parameters based on household inter- dices; then estimate demand equations on the assump- view data from Bogota and Cali, Colombia, are given. A tion that residents of each neighborhood face only the comparison of parameter values with those obtained prices in their own neighborhood (King 1975). from North American data sets shows that the Co- 3. Allow intrametropolitan prices to vary by indi- lombian demand elasticities are generally comparable in vidual dwelling unit; estimate a dwelling unit price in- magnitude to those from the United States. Second, an dex with the use of a production function for housing approach, derived from residential location theory, that and varying input prices; estimate demand equations on stratifies people by place of work is employed to repre- the assumption that occupants of each dwelling unit sent housing price variation in the demand equations. face only the price of their own dwelling unit (Polinsky Finally, a simple exercise illustrates the magnitude of and Elwood 1979). bias of the income elasticity of demand that can result 4. Allow the marginal cost of attributes to differ from incorrect data aggregation techniques. Correctly within a metropolitan area; estimate a nonlinear he- aggregated data produce income elasticity estimates donic price index and use the first derivative of the index that are similar to those obtained from disaggregate or with respect to specific attributes as the price term in a micro data. demand equation for the attribute (Witte, Sumka, and Erekson 1979). The Price Term in Housing Demand Each approach has potential shortcomings. Omission Equations of price variation, as in (1), can bias other demand equation parameters if the omitted price term is corre- Estimating demand equations for housing from cross- lated with included variables. The assumption that sectional data presents many challenges, but measuring households face only their neighborhood or dwelling the variation in the unit price of housing is probably one unit prices, as in (2) and (3), may lead to a fundamental of the greatest difficulties. Data sets typically report the misstatement of the price variation in the sample if total expenditure on housing rather than a unit price households are not limited in their choices to specific and the quantity of housing. Hence the unit price must neighborhoods or dwelling units. If all purchasers face be inferred by relating variations in expenditure to varia- all prices, the price "chosen" may reflect the impact of tions in quantity. Moreover, housing is inherently mul- other household characteristics. Price variations based tidimensional; it includes attributes of size, dwelling on neighborhood or dwelling unit require that market Housing Demnand in the Developing-Country Metropolis 137 segmentation based on those dimensions be justified. that has the lowest housing price index. Urban econo- Estimation of demand equations for specific attributes of mists have long argued that a metropolitan area with housing, as in (4), may not be relevant if we are really multiple workplaces and a price gradient for housing interested in the demand for housing as a composite must have differential wage levels across workplaces for good. households to be in equilibrium (Moses 1972). Accord- A relatively simple application of residential location ingly, workplaces can have different housing prices, but theory suggests an alternative way of incorporating price they then must also have compensating differentials in variation into a demand equation for housing as a com- wages to keep households in equilibrium. The existence posite good. Simple models of residential location of wage gradients across workplaces thus becomes a theory are essentially based on the precepts of cost necessary condition for the workplace-based housing minimization. A worker surveys the housing market price variation approach taken here.' from his workplace,j. He typically observes that hous- ing prices, R, decline with distance, d, from his work- place in at least one direction, but that travel costs, t, increase with distance from his workplace. For any given Housing Demand and Workplace-Based amount of housing, H, he faces a total expenditure on Price Variation housing, Z, composed of a housing expenditure plus a transport expenditure, In developing a workplace-based price index for hous- (11-1) Z= R (d)H + t (d). ing, there are two possible formulations for the demand system, each of which uses a different definition of the For quantity H0 the worker can solve for the least-cost price of housing. Differing definitions alter the specifica- distance by taking derivatives tion of the demand equations. In one formulation the (11-2) Zjl ' Rj'(d)HO + t '(d) = O price of housing is based only on the housing expendi- (*1-2) ' R1'(d)H0 + t~'(d) ture and does not include the travel expenditure. In this and solving the expression for dj*, the optimal distance or case the budget constraint is written location for quantity Ho and workplacej. This least-cost (11-4) Y=PHH+PV+t(d) distance can be substituted back into equation 11-1 to ( calculate the minimum total expenditure for quantity where Y is income, PH the price of housing, and P, the HO, price of composite commodity V. In this formulation the (11-3) Zj* = Rj (dj*) Ho 0+ tj (dj*) . travel expenditure, t (d), is included in the income con- ('1-3) = R1(dfl H0 straint, and the derived demand equation is Consider carrying out this exercise for different work- (11-5) H=f{PH, [Y-t(d)]}. places in a metropolitan area. The decline of housing prices with distance, R,(d), differs systematically across That is, travel costs have to be subtracted from income workplaces and probably shows steep rates of decline in the demand equation. If travel costs are an unknown with distance for centrally located workplaces and grad- function of distance, d, then d is included in the demand ual rates of decline for peripheral workplaces. Travel equation as a separate variable. costs per unit distance may also differ by workplace, but In the second possible formulation the price of hous- in ways that may be difficult to generalize. For example, ing is the so-called gross price and is based on the transit speeds may be higher but transit headways housing expenditure plus the travel expenditure. In this longer at peripheral locations as compared with a cen- case the budget constraint is tral location. As the workplace varies, however, so does (116) Y=ZH+PV the optimal housing and travel expenditure required for ( - HH - housing quantity Ho. This variation in expense by work- where ZH is the gross price term. The travel cost does place for a given quantity of housing is used as a measure not enter separately into the budget constraint, and the of price variation in the housing demand equations distance term will not appear in the demand equations. estimated here. A price index is estimated for each work- To implement this second approach, however, one must place zone. Households whose heads work at a particular be able to specify a priori the travel cost function, which workplace zone face the same housing price index, is a combination of out-of-pocket cost and the opportu- households with heads at another workplace face the nity cost of travel time. Since insufficient information is price index at their workplace, and so forth. Prices will available for Bogota and Cali to permit specification of vary by workplaces. the travel cost function with confidence, the first If housing prices vary by workplace, it is worth asking approach has been implemented here. The estimated why all workers do not try to obtain jobs at the workplace demand equations therefore include distance to the 138 Gregory K. Ingram workplace, as in equation 11-5, and the workplace-based The cost of a standardized unit is used to formulate a price term is based on housing expenditure only. workplace price index by choosing workplace 1 as a The relevant housing expenditure that is used to numeraire and calculating a price index define a price index for a given workplace is the efficient, or optimal, expenditure implicit in the solution of equa- (11-10) Rj tions 11-2 and 11-3. Corresponding to each quantity of R, housing, H, is an optimal location or optimal distance, The households in the sample also have associated with d *, and an optimal expenditure, R (d *)H. If households them C household characteristics, HCc, that affect employ the kind of locational calculus embodied in res- household demand for housing. These characteristics of idential location theory, the choice made by households the households and the distance from home to work, dij, with a head employed at a particular workplace is at or are used in a demand equation whose dependent variable near the optimal location for that workplace, and their is housing expenditure divided by the price index in housing expenditure approximates the optimal expendi- equation 11-10, or a quantity index of housing. The ture for their workplace and housing quantity.2 The demand equation is of the form relation between housing expenditures and housing R quantity for a given workplace can be captured by re- (11-11) '= f(H. HCij dij) gressing the observed housing expenditure on measures "i of housing quantity for households whose heads work in and is estimated over the sample of allM households as a the same work zone. The relation between housing ex- single pooled demand function. In equation 11-11 both penditure and housing quantity can then be used to linear and double log specifications are used. formulate a price index for the given workplace. This procedure can be repeated to yield a price index for each workplace, and these workplace-specific price indexes The Setting and the Data can then be used as a price term in demand equations for housing as a composite good. The household interview data used to implement the The specific procedure that has been implemented in housing demand procedure just outlined are from Bo- this chapter can be summarized as follows. The sample gotA and Cali, Colombia. The principal data set used was contains M households whose household heads have collected in 1978 and covers both owners and renters, jobs located at one of J work zones, and there are Nj for whom equations are estimated separately, in Bogota households associated with workplacej. For household i and in Cali. A second data set is available for Bogota in (i = 1 toNj) at workplacej, the monthly expenditure on 1972, but data for renters only can be used to estimate housing (or the dwelling unit value), Rij, and a set of K the demand for housing in 1972. dwelling unit characteristics, Xijk, are known. For each In 1978 BogotA had a population of roughly 3.5 mil- of the J work zones an estimation is made of the equa- lion and Cali a population of roughly 1.1 million. Both tion cities have experienced rapid rates of population growth K in the past (BogotA's population in 1972 was 2.8 mil- (11-7) Ruj = kY BjkXijk lion), but current population growth rates are moderat- ing in both cities. Per capita income in 1978 was about by regressing housing expenditure on the measure of $800 a year in both places. The cities differ significantly dwelling characteristics, and J sets of parameters which in climate because of their different altitudes: BogotA is indicate how the cost of housing attributes varies by 8,000 feet above sea level and has temperate weather workplace are obtained. A representative dwelling unit with cool nights; Cali, at 3,000 feet above sea level, is in the housing market is then defined as the unit that semitropical and warmer than BogotA. Differences in has the samplewide average amount of each dwelling size and climate may explain some of the differences in unit characteristic, where the average quantity is housing demand in the two cities. - 1 Nj To implement the workplace-derived price indexes it (11-8) X* = M ' k X-jk * was necessary to divide the two cities into a number of (11-8) Mi = 1 work zones the boundaries of which are arbitrary but are The dwelling with attributes Xk then becomes the stan- based on considerations that include compactness, re- dard unit-the equivalent for housing of the standard- spect for significant internal divisions, and a require- ized market basket. For each workplace the estimated ment that there be an adequate number of observations parameters in equation 11-7 are used to calculate the in each work zone. The same work zone system was used cost of the standard unit, in BogotA in 1972 and 1978 and for renters and owners (11-9) Rji = ;BIkXk. in each city. Tabulations of residence and workplace by k Housmg Demand in the Developing-Country Metropolis 139 annular ring and radial sector indicate a high degree of ber of rooms, presence of garbage collection, distance to association between place of work of the household head nearest bus line, type of structure, condition of unit, and place of residence of the household. Empirical presence of nonresidential use, and presence of a public analyses indicate that although the workplace of second- or private phone. Current prices are used in both time ary workers may have a slight influence on a household's periods. The consumer price index had approximately residential location, the workplace of the household tripled, from 47 in 1972 to 150 in 1978. head is clearly a dominant determinant (Pineda 1981). The coefficients from the hedonic price equations are Average commute lengths in kilometers for each work most comparable for 1978. In 1978 there are equations zone and tenure type in both cities differ by up to a factor for thirteen Bogota and eight Cali work zones and for of 3. In both cities commute lengths are long for central- two tenure types, a total of forty-two equations. The only ly located workplaces and for workplaces located along variable that always has the correct sign in all forty-two the mountains. equations is dwelling unit area. Access to a phone, exclu- sive bath and kitchen facilities, and garbage collection also perform well; they have the expected sign thirty-six, The Hedonic Price Equations thirty-seven, and thirty-two times, respectively. The number of blocks to a bus is positive only half of the Separate hedonic equations were estimated for each time; it is possible that there is some disamenity associ- work zone and tenure type in Bogota and Cali in 1978. ated with being too close to a bus route. Lot area does For 1972 in Bogota a hedonic equation was estimated not perform well in the hedonic equations, and it does only for renters because no data were available for the very poorly in Cali, where owners, in particular, do not value of owner-occupied units in the 1972 sample. For seem to value larger lot size. The hedonic equations for renters the dependent variable is the monthly rent; for the 1972 Bogota renters are similar to those for 1978 in owners the dependent variable is the value of the dwell- that the measure of interior space, the number of rooms, ing unit in thousands of pesos. has a positive effect on rent.3 Because the 1978 data were all collected in the same A measure of the explanatory power of the hedonic survey with the same questionnaire, it is possible to use price equations is shown in table 11-1, which summa- the same specification for the four sets of 1978 equa- rizes the explanatory power of the regression equations tions. In the 1978 equations the independent variables and the workplace stratification in an analysis of included the dwelling unit area in square meters, variance framework. Overall, the analysis explains from DUAREA, the lot area in square meters, LOTAREA, the 45 to 69 percent of the variation in housing prices; the number of blocks to the nearest bus line, BLKTOBUS, a equations have much more explanatory power than does dummy variable equal to 1 if the residence had a private the workplace stratification. Interestingly, the work- or public phone, DPHONACSS, a dummy variable equal place stratification has more explanatory power for to 1 if the dwelling unit had its own nonshared kitchen owner-occupied units than for renter-occupied units. and bathroom facilities, DEXCLUSE, and a dummy vari- This is consistent with the empirical regularity that able equal to 1 if the dwelling unit had its garbage picked owner-occupied units have steeper price gradients in up by municipal authorities, DGARBCOL. It is interest- urban areas than do renter-occupied units. Hence, ing to note the similarities and differences between workplace location matters more in the owner market tenure classes and cities in the independent variables. than in the renter market. Renters in Bogota and Cali, for example, have similar- size units on similar-size lots, but Bogota renters have Table 11-1. Analysis of Variance: more phones, while Cali renters have better garbage Hedonic Pice Equations collection. Bogota owners have larger, more expensive Percent of variation explained homes on larger lots than do Cali owners. The most by striking differences between renters and owners are in Work zone the average area of the unit and the proportion of units Data stratification Equations Total having exclusive bath and kitchen facilities; owners are better housed than renters. Finally, there is more Bogota variability in the average dependent variable across work 1978 renters 2.5 47.6 50.1 zones than there seems to be in the average independent 1978 owners 8.7 36.4 45.1 variables, which indicates that prices do vary. CaI The independent variables used in the 1972 equations Cal differ from those used in 1978 because the questionnaire 1978 owners 8.0 60.9 68.9 was quite different. The variables are building age, num- 140 Gregory K Ingram Map 11-1. Bogota: Workplace Price Indexes, independent variables are monthly household income (a Thirteen Work Zones, 1972 and 1976 measure of current income) in pesos, the price index described above, and the air distance from home to Note: Data shown for each zone are, in order, work, in meters. Three additional household character- 1972 rent index S istics are included in the demand equations: a dummy 1978 rent index variable for the sex of the household head (1 = male); 1978 value index family size (the number of persons in the household); Zone I 1.0 and the age of the household head. The hypothesis is that these three characteristics capture differences in J(3 <> 0.82 r) taste (sex of household head), in the need for housing 2 > 0.84 (family size), and in assets or wealth (age of the house- hold head). Two functional forms, double log and linear, are estimated. In the linear specifications squared terms no v< 1.23 /for family size and the age of the head are entered to < / \ 0.99 , capture nonlinearity in the effects of those variables. 0. <;~\/ zon; 72 <1.36 tFive sets of equations are estimated; each set is a Zoe \1 0.80 1 .n different combination of year, tenure choice, and city. A comparison of mean values across the five samples shows that renters have younger household heads, 0.73 - Zone 4 smaller families, and lower incomes than owners. Differ- f \_~ zone 10 \ 075 Z one 3 1 -15 t 077 \tsot<\l 0.8 1 .07 Map 11-2. Cali Workplace Price Indexes, Zone 1 Eight Work Zones, 1978 + () ~~~Zo°ne6 A0oe > +0 76 / A 1. Note: Data shown for each zone are, in order, zone 9 oo 5 xo 0.62 1978 rent index 0874 X C 1978 value index Zone 1 = 1.0 Zone 3 0.90 0.99 one 4 0.104 0.66 Zone 7 Zne 20.75 The standardized rents and values were obtained by 1.15 068 plugging the average renter and owner unit characteris- 1.0 1.03 tics for Bogota and Cali into their respective workplace 1 0.76 hedonic equations. For use in the demand equations these rents and values are transformed into spatial price Zone 6 0.98 indexes by dividing through by the relevant rent or value 0.64 for work zone 1, the central business district. The result- ing normalized price indexes are displayed for Bogota in map 11-1 and for Cali in map 11-2. There obviously is Zn I variation across work zones in these price indexes. In 0.86 both Bogota and Cali there is more variation in the price 0.64 index for owners (the range covers a factor of 2) than for renters. The Housing Demand Equations The dependent variable in the demand equations is the monthly rent or value divided by the workplace- specific price index, as shown in equation 11-11. The Housing Demand in the Developing-Country Metropolis 141 ences between Bogota and Cali are slight except for vary among the samples and specifications shown, but income; Bogota owners have much higher average in- they display a consistent and stable pattern for most of comes than do Cali owners, and Bogota renters have the variables. All income elasticities are less than one, average incomes similar to those of Cali renters. 1978 and at the sample mean they lie in a narrow range of 0.6 Bogota renters had smaller families and younger house- to 0.8 except for the equations for Cali renters. The hold heads than did 1972 Bogota renters. elasticity of the sex of the head is always negative and The demand equations perform well with R2 statistics small (in the interval 0 to - 0.2). Family size elasticities that range from 0.25 to 0.6. Income is by far the most show an interesting pattern; they are negative for own- important explanatory variable. Age of the head and ers and are usually positive for renters. Since renter- family size are usually significant; sex of the head is occupied units are usually smaller than owner-occupied usually not significant, but it always has a negative sign. units, it appears that space is a binding constraint for The housing price index is significant in two of the five renters, and larger renter families obtain more housing. samples, and it always has the correct sign. Distance Owner-occupants seem to be able to reduce the quantity from home to work is significant in four of the five of housing demand as family size increases because they samples and has the correct sign in nine of the ten have larger units on the average and the quantity of equations. housing is not constrained by family size. Age of the A summary of results from the fully specified demand household head has a consistently positive demand elas- equations, in the form of elasticities for each indepen- ticity when evaluated at the sample mean. By using the dent variable, is displayed in table 11-2 for renters and linear demand equation with the squared term for age of table 11-3 for owners. The elasticities are calculated in the head, it is possible to calculate the age at which the linear equations with the use of the mean value of housing demand is at a maximum. This is consistently each independent variable except income. The linear ages 50-57 except for Bogota owners, for whom it in- elasticities are shown for the first, second, and third creases throughout the relevant range. The price elastic- quartiles of each sample's income distribution. In each ity of demand lies consistently in the interval 0 to - 1.0 case the sample mean and the seventy-fifth percentile of in absolute value and becomes absolutely quite small for the income distribution are essentially identical. some of the linear specifications. Finally, the distance The magnitudes of the various elasticities obviously elasticity is almost consistently negative and quite small. Table 11-2. Demand Elasticities at Various Income Levels, Renters Home- Income percentile Sex of Family Age of to-work and levela Income head size head Price distance 1972, linear, Bogota 25 (1,000) 0.32 - 0.16 0.30 0.23 - 0.91 - 0.05 50 (1,700) 0.45 -0.13 0.25 0.19 - 0.75 - 0.04 75 (3,079)b 0.59 -0.09 0.18 0.14 -0.55 -0.04 1972, log-log, Bogota All 0.77 -0.14 0.14 0.12 -0.70 -0.06 1978, linear, BogotS 25 (3,500) 0.55 -0.03 - 0.24 0.95 - 0.17 - 0.23 50 (7,100) 0.71 -0.02 - 0.16 0.61 - 0.11 - 0.15 75 (11,260)b 0.80 -0.003 -0 .11 0.43 -0.08 - 0.10 1978, log-log, Bogota All 0.72 - 0.07 0.10 0.07 - 0.28 - 0.06 1978, linear, Cali 25 (3,500) 0.05 - 0.01 0.48 0.62 - 0.34 - 0.16 50 (7,300) 0.10 -0.01 0.46 0.59 -0.32 -0.15 75 (12,829)b 0.16 -0.01 0.42 0.55 -0.30 -0.14 1978, log-log, Cali All 0.47 - 0.20 0.36 0.43 -0.48 -0.03 a. Income levels are in parentheses. b. Sample mean. 142 Gregory K Ingram Table 11-3. Demand Elasticities at Various Income Levels, Otvners Home- Income percentile Sex of Family Age of to-work and level' Income head size head Price distance 1978, linear, Bogota 25 (6,000) 0.33 - 0.03 -0.34 0.66 -0.31 0.02 50 (10,900) 0.47 -0.02 -0.27 0.52 -0.24 0.01 75 (17,942)b 0.60 -0.02 -0.21 0.40 -0.19 0.01 1978, log-log, BogotA All 0.78 - 0.09 -0.25 0.25 - 0.44 -0.02 1978, linear, Cali 25 (5,000) 0.39 -0.06 -0.57 0.53 - 0.27 -0.06 50 (8,800) 0.53 -0.05 -0.44 0.18 -0.21 -0.05 75 (13,841) 0.64 -0.04 -0.34 0.13 -0.16 -0.04 1978, log-log, Cali All 0.76 -0.06 -0.30 0.08 -0.33 - 0.02 a. Income levels are in parentheses. b. Sample mean. Table 11-4. Range of Housing Demand Elasticities in Various Countries (from Household Observations) Elasticity of housing demand with respect to Current Family Age of Sex of Country income Price size head head (I =male) Renters Colombia 0.2 to 0.8 -0.1 to -0.7 -0.1 to 0.4 0.1 to 0.6 -0.01 to -0.2 United States 0.1 to 0.4 -0.2 to -0.7 ? ? Consistently negative Korea, Rep. of 0.12 -0.06 to 0.03 0.15 to 0.25 - - Owners Colombia 0.6 to 0.8 -0.15 to -0.40 -0.2to -0.35 0.1 to 0.4 -0.02 to -0.1 United States 0.2 to 0.5 - 0.5 to - 0.6 ? ? Negative Korea, Rep. of 0.21 - 0.05 to 0.07 -0.02 to 0.15 ? Not conclusive. - Not reported. Sources: United States: Mayo (1981); Korea: Follain, Lim, and Renaud (1980). Table 11-4 summarizes the range of demand elastici- States. There are three possible explanations for this ties obtained in Cali and Bogota and compares them result. First, female-headed households may have with estimates obtained from household surveys in the stronger preferences for housing than male-headed United States and Korea. The general pattern of results households. Second, female-headed households may be is quite similar for the United States and Colombia; both discriminated against and face higher prices, which countries differ somewhat from Korea. The Colombian could produce larger expenditures on housing, and income elasticities are somewhat higher than those those larger expenditures could show up as a preference obtained in the United States, whereas the Colombian for larger quantities in the demand equations for rent- price elasticities may be lower. The elasticities of hous- ers. The discrimination hypothesis is, however, uncon- ing demand with respect to family size and age of the vincing for owner-occupants. Third, female household head in Colombia cannot be compared with data from heads have shorter commuting distances than male the United States but are somewhat similar to the Ko- household heads and may therefore systematically pay rean estimates. Finally, the effect of the sex of the house- higher prices for housing, since rents decrease with hold head, although usually statistically insignificant in distance from the center of the city. The demand equa- Colombia, is also always negative, as in the United tions used should account for this difference, however, Housing Demand in the Developing-Country Metropolis 143 because distance is included. Accordingly, the first ex- hand-held calculator. The equations yielded by this exer- planation, based on preference differences, may be the cise are shown in table 11-5, and the resulting income most plausible. elasticities are compared with those from the disaggre- To investigate the effect of distance on the coefficient gated, fully specified equations in table 11-6. The aggre- for the sex of the head and to see how sensitive the other gate estimates each differ by less than 20 percent from parameters are to both the price and distance terms, the the disaggregated log-log estimates, and in four of five housing demand equations were estimated without cases the aggregate log-log estimates lie between the those terms. Omission of the price and distance terms linear and log-log disaggregated estimates. It is obvious tends to reduce the income coefficient very slightly, that aggregate-based estimates of the income elasticities often only in the third significant digit. The family size of the expenditure for housing could be a good approx- effects are also affected only minimally by the omission imation for the income elasticity of demand for housing of the two terms. The coefficients for sex and age of head in the samples used hlere. do change considerably in percentage terms, however; It is important to note that the aggregate estimates this seems to be largely a consequence of the omission of obtained are sensitive to the way in which the under- the distance term. Female-headed households live closer lying micro data are aggregated. An experiment that to the head's workplace than do male-headed house- illustrates this was performed with the 1972 sample of holds, as do households with older heads compared with renters. The sample was aggregated to the level of sixty- households with younger heads. In general, however, three zones for the city of Bogota, and average rents and the parameter estimates for the included variables are incomes were calculated for each zone. A hand-held stable with respect to the omission of the price and calculator was then used to calculate a log-log regres- distance terms. These exercises suggest that neither the housing prices as specified in these demand questions nor the Table 11-5. Housing Demand Equations distance from home to work are collinear with house- from Aggregate Data hold income. Indeed, in Bogota and Cali, as in many Sample B, B, R2 other cities, the use of micro data dramatically reduces problems of multicollinearity in the estimation of hous- 1972 phase II renter 2.92 0.71 0.99 ing demand equations. 1978 Bogota renter 1.54 0.79 0.99 1978 Cali renter 12.38 0.55 0.97 1978 Bogota owner 9.11 0.67 0.99 Aggregate Estimates of Income Elasticities 1978 Cali owner 7.81 0.66 0.97 Note: The equation is of the form R = Bo - BY, where R is rent and All of the parameter estimates that have been pre- Yis income. sented so far have been obtained from computer-based multivariate regressions that use individual households Table 11-6. Aggregate and Disaggregated as observations. In many situations it may not be possi- Income Elasticities of Housing Degand ble to gain access to individual household records be- cause of confidentiality restrictions, while in other situa- Sample and tions lack of time or of adequate computer facilities may make parameter estimation with micro data impossible. 1972 Bogota renter This section briefly investigates the adequacy of the Log-log 0.71 0.77 parameter estimates that could be made from published Linear - 0.59 aggregated data. The focus is on the estimation of the 1978 Bogota renter income elasticity of the demand for housing because Log-log 0.79 0.72 that parameter is often of interest in both the design and Linear - 0.80 1978 Cali renter the evaluation of housing programs, policies, and Log-log 0.55 0.47 projects. Linear - 0.16 Each of the five samples analyzed above is now sum- 1978 Bogota owner marized in a matrix dimensioned by rent or value and by Log-log 0.67 0.78 income. Eight income categories were defined for the Linear - 0.60 1978 data and nine for the 1972 data. The average rent or 1978 Cali owner value was calculated for each income category; this Log-log 0.66 0.76 average was then regressed on the midpoints of the Linear - 0.64 income categories in a log-log specification using a -Not applicable. 144 Gregory K. Ingram sion of average zonal rent on average zonal income, nous to the local economy and reflects, among other using all sixty-three observations. The resulting income things, local income levels. Perhaps we should be more elasticity, 0.95, was substantially higher than the 0.71 surprised at the similarities between Bogota and Cali, estimate that was obtained with nine observations from which have markedly different climates. the correctly aggregated sample. A second experiment Simple experiments involving the aggregation of the was then run on the 1972 Bogota data. For this experi- household survey data used to obtain micro data esti- ment the data in the rent-income matrix were incor- mates suggest that income elasticity estimates on the rectly aggregated by calculating the average income for basis of correctly aggregated data can be good proxies for each rent category and regressing the rent category estimates on the basis of fully specified models that use midpoints on the mean incomes. This rent-stratified household observations. Estimates on the basis of incor- approach yielded an income elasticity estimate of 1.36, rectly aggregated micro data, however, can produce nearly twice the 0.71 obtained with the use of an in- estimates of the income elasticity of demand that are come-stratified aggregation procedure. It is obvious that badly biased. the aggregation bias in estimates of income elasticities can be very large, but that correctly aggregated data can give useful results. Notes 1. Preliminary empirical work indicates that a wage gra- Conclusion dient with a peak in the central business district does exist in Bogota. This paper has described and implemented a two-step 2. Equation 11-3 can be solved for the expansion path of expenditures as the quantity of housing increases. estimation procedure for incorporating price variation 3. Parameter estimates are available from the author. in the estimation of demand equations for housing, using household survey data from Bogota and Cali, Co- lombia. The demand equations estimated with this pro- Bibliography cedure give significant results for the income elasticity of the demand for housing; estimates of the income Follain, J., G. C. Lim, and B. Renaud. 1980. "The Demand for elasticity generally lie in the upper end of the 0.2-0.8 Housing in Developing Countries: The Case of Korea."Jour- range. Although the price term in the demand equations nal of Urban Economics, vol. 7, no. 3 (May), pp. 315-36. gave less significant results, the price elasticity of de- King, Thomas. 1975. "The Demand for Housing: Integrating mand appears to be between 0 and - 1. There is, howev- the Roles of Journey-to-Work, Neighborhood Quality, and er, greater uncertainty about the magnitude of the price Prices." In N. Terleckyj, ed., Household Production and elasticity than about the magnitude of the income elas- Consumption, pp. 451-83. New York: Columbia University ticity. Other household characteristics involved in the Press. demand equations have low demand elasticities, typical- Mayo, Stephen K. 1981. "Theory and Estimation in the Eco- ly less than 0.5 in absolute magnitude. The age of the nomics of Housing Demand."JournalofUrban Economics, head has a positive elasticity over most of its range, while vol. 10, no. 1 (July), pp. 95116. family size usually has a positive elasticity for renters Moses, Leon N. 1972. "Toward a Theory of Intra-Urban Wage and a negative elasticity for owners. The demand equa- Differentials and Their Influence on Travel Patterns." Pa- andanegativnssugeeastiaty foraowners.aTheddemd equseho pers and Proceedings of the Regional Science Association. tions suggest that female-headed households consume .. more housing than male-headed households, but this Muth, Richard F. 1969. Cities and Housing. Chicago: Univer- result is rarely statistically significant. Distance from sity of Chicago Press. home to work is entered into the demand equations as Pineda, Jos6 Fernando. 1981. "Residential Location Decisions an adustmnt t incme, ut i undubtely aso rpre- of Multiple Worker Households in Bogota, Colombia." Pre- an adjustment to income, but it undoubtedly also repre- sented at annual meetings of the Eastern Economic Associa- sents price variation within the workplace strata, which tion, Philadelphia, Pa., April. Processed. are used as the main representation of price variation. Polinsky, A. Mitchell, and David M. Elwood. 1979. "An Empir- The distance elasticity is small-less than - 0.2-and is ical Reconciliation of Micro and Grouped Estimates of the almost always negative. Demand for Housing." Review of Economics andStatistics, Comparisons of elasticity estimates with those vol. 61, no. 2 (May), pp. 199-205. obtained from U.S. data sets indicate that the range of Witte, Ann D., Howard J. Sumka, and 0. Homer Erekson. the Colombian estimates generally overlaps the range of 1979. "An Estimate of a Structural Hedonic Price Model of the U.S. estimates. This similarity of values may seem the Housing Market: An Application of Rosen's Theory of surprising at first but is much less so on reflection. Implicit Markets." Econometrica, vol. 47, no. 5 (Septem- Housing is a nontraded good, and its price is endoge- ber), pp. 1151-73. 12 Urban Transport Policy: Colombia Alvaro Pach6n O ur knowledge of urban transport in Colombia explained by the high percentage of trips made on public c comes in great part from transport and urban fi- transport and the low rate of private motorization (see nance studies made during the past thirteen years in the Pach6n 1981c; Westin 1980). The main elements of what big cities and especially in Bogota. Discussions have is known about collective transport in Colombia are centered on two main themes: public versus private summarized below. transport, and choices in public transport. State Intervention in Collective Transport Operations The state has recognized collective transport as a public service and a basic need. Inability to charge di- The less efficient use that private vehicles make of rectly for road use leads to excessive use of private road space and fuel per passenger has been identified as vehicles and inefficient use of road space, and additional one of the causes of the transport problem (see World inefficiencies are created in a system of free competition Bank 1975, pp. 84-85; Urrutia 1981, pp. 12-13). To when drivers do not consider the negative effects of their improve resource allocation, two kinds of measures have behavior on other vehicles on a congested road. Collec- been suggested: taxes and subsidies to shape the trans- tive transport can reduce some of these problems. As for port market, and quantitative restrictions to limit use of transport as a basic need, under free competition, low- the roads. income consumers may not be served when the cost of A proposal that relies on market mechanisms is the providing service is above what they can afford, and state use of a fuel tax to control urban congestion (Republic of intervention may be warranted. Intervention may be Colombia 1973, pp. 99-111). McClure (1974) has pro- direct, through production or pricing of the basic need, posed a parking tax. Although the government has not or indirect, through income transfers. As high-income accepted either proposal, it has adopted some measures classes seldom use buses, direct intervention is likely to that clearly favor public over private transport, includ- use less resources to meet the basic need objective and ing sales and import tax exemptions for public transport hence is preferable to indirect intervention. vehicles, bus fares that are set below average cost, and lower fuel prices for commercial vehicles. The main restriction designed to favor collective transport is the Organization of Collective Transport prohibition of private vehicles on Carrera 10, one of There are two types of ownership of collective trans- Bogota's principal streets. (In contrast, the local trans- port firms: ownership of the vehicles by the municipal port authority in Cali has set restrictions on bus access government-the predominant arrangement in the to the traditional center.) United States-and ownership by private organizations, The considerable amount of investigation that has as in Argentina, Brazil, Chile, and Colombia. Colombia been devoted to the subject of collective transport is has a basic system of affiliating companies which coexist 145 146 Alvaro Pachon with state transport companies and cooperatives. In Nacional de Planeaci6n-Instituto Nacional de Trans- Bogota's system of affiliating companies the state porte (INTRA) (1974). By means of a route simulation assigns the routes and gives some rights to the private model, these studies bring out the savings, in operating company. In some cases the company owns some of the costs, of a centralized collective transport system. buses, but generally bus owners affiliate with a company Economic evaluation of the two types of organization which distributes the routes and charges its associates a requires a simultaneous analysis of the two types of rolling charge for use of its routes. The managers of this efficiency that takes into account the variations state type of organization are the strongest group in the ownership introduces in the operating costs and consid- transport system because they maintain a close relation ers the benefits and costs associated with trip duration with the official sector, through which they obtain oper- and waiting and transfer times. In addition, political ating licenses and route authorizations. considerations are relevant. Which type of organization In the state companies, the state administers and better promotes the satisfaction of basic needs? Which owns the transport equipment, as in the case of the type is more reliable and less affected by strikes? Empresa Distrital de Buses de Bogota (District Bus (Although strikes are prohibited in government services, Company of Bogota). This company's importance has slowdowns are often used for the same purposes.) decreased; in 1980 it accounted for less than 1 percent of public transport buses and busetas (minibuses). Coop- The Route System eratives, which own their transport equipment, with the associated drivers having shares, predominate in The route systems in Colombia and especially in some cities, such as Cali, and occupy a second place in Bogota have been frequently studied. Cifuentes (1978) others, including Bogota. notes that authority to assign routes has been delegated In 1980 Bogota had thirty-eight public transport com- by INTRA to the mayoral offices in the principal cities. The panies, of which twelve were cooperatives, one was state- resulting separation of functions has made control of the owned, and twenty-five were limited liability companies public transport system more difficult. (affiliating companies). In Cali in 1979 there were 1,426 To assign new routes, or to extend existing routes, the buses distributed among sixteen private companies with Administrative Department of Transit and Transporta- forty-two routes and 181 busetas distributed among tion (DATT) studies the potential demand and opens a three companies with seven assigned routes. There has tender to assign the new route. The tender specifies the been considerable debate about the organization of pub- number of vehicles required and the frequency of service lic transport. Defenders of a public collective transport during peak and nonpeak hours. The assignment pro- system have pointed out the inefficiencies in resource cess has been much criticized. It has been argued that allocation associated with the existing system in Co- the system does not respond to the real needs of the lombia, while the defenders of the present system em- population because the decisions respond to pressure by phasize the low costs of private companies. In Leiben- the Juntas de Acci6n Comunal (community action stein's terminology (1966), the defenders of a statal boards) of the districts and do not take into considera- system focus on efficient resource allocation, while the tion the needs of the whole population. It has also been defenders of the present system focus on the inherent argued that the prevailing system of routes and route X-efficiency of private property. Interconsult Ltda. assignments produces high levels of congestion in the (1970) and Urrutia (1981) have documented the existing central area because all the companies prefer to serve system's low operating costs, the good maintenance areas where demand levels are high. The existence of level of the vehicles, and the system's ability to adapt to parallel routes and the information problem caused by changes in the demand for transport. A 1980 study by the great number of routes have also been criticized. To the Administrative Department of District Planning remedy some of these problems the Interconsult Ltda. shows the high levels of X-inefficiency of Bogota's public and INTRA studies proposed the consolidation of routes company. Walters and Feibel (1980) and Urrutia (1981) on the basis of the results of a simulation model. also show that in other parts of the world private com- It has, however, also been established that the existing panies have lower cost levels and are more X-efficient route assignment system in Colombia produces a net- than public companies. As an exception, Nelson (1972) work with wide coverage that has adapted to changes in shows that in the United States in 1960 and 1968 public trip demand patterns. The efficiency with which owners, companies had costs 10 percent lower than those of drivers, and users process the existing information to private companies. obtain a satisfactory allocation has been noted. The inefficiency of resource allocation, especially as Discussion of the route system in Bogota leads to concerns congestion, has been documented by, among some important conclusions. First, a series of incentives others, Interconsult Ltda. (1970) and Departamento is bringing about a satisfactory solution. The urban Urban Transport Policy: Colombia 147 transport market is processing a great deal of informa- To balance supply and demand for urban transport, tion at a low cost. If there were no externalities, the the state has enunciated policies which are sometimes result produced by the market would be the same as contradictory. For example, it has established a financial would be obtained in a model for the whole transport intermediary, the Financial Transportation Corporation system. Second, given the existence of such externalities (cF-r), with credit lines for financing body work and as congestion and buses' stopping on demand, the pres- chassis for buses and busetas. This credit has been subsi- ent system of decentralized decisionmaking may be dized to stimulate investment in public service vehicles faulty. Third, since there is no charge for road use in the and to facilitate the purchase of vehicles by the drivers. central area, the income from scarce road space is trans- The subsidy, which amounts to nearly a fifth of the ferred to the private sector, probably to the affiliating vehicle's value, has undoubtedly made investment in companies. Fourth, although the existing data allow the public vehicles more attractive. use of simulation models developed for Colombia, it is The state also wants to maintain an adequate trans- probable that more effort is needed to collect and ana- port supply and assure an adequate income yield for new Jyze data and to refine the models if their results are vehicles. Differential fares were established during the needed to assign a route system. Fifth, to ensure good late 1960s according to the age of the vehicle, with service to areas with difficult topographic conditions and higher fares for newer vehicles. Since this system did not high operating costs, it is necessary to specify differen- work, a differential subsidy by age was tried. The existing tial fares or subsidies for routes that serve those areas. subsidy system does not fulfill the initial objective of Sixth, if the state established a fee for the right to operate favoring investment in new buses. routes that serve the central areas, resource allocation Moreover, the government has established a license- could be improved and funds could be generated to known in Bogota as a nota opci6n-for new public provide subsidized service to the low-income districts service vehicles. This entry restriction has artificially that have unfavorable cost conditions. limited the size of the vehicle stock and has created some income for those who already own public service ve- Public Service Vehicles hicles. Also, because in some cities the procedure for granting licenses is stricter for small vehicles, the mix of The initial investigations by Interconsult Ltda. and vehicles in the stock is artificially distorted. mNTRA considered it desirable to define the technical char- acteristics that would assure more efficient and comfort- able service. It was thought that metropolitan buses like Remuneration of Drivers those in the United States should be used and that the The public companies pay drivers a monthly salary for school-bus type of vehicle should be taken out of circula- a normal eight-hour day; the private companies pay a tion. The success that microbuses and busetas were commission according to the number of passengers having at the end of the 1960s was acknowledged, transported. The private companies' system has been however, and was attributed to their greater speed, criticized because it may lead drivers to compete to pick privacy, and comfort. up passengers without regard to traffic laws. It has also In Colombia as in other developing countries it was been pointed out that the driver is being exploited and soon recognized that although the metropolitan bus had has long working days and unfavorable working condi- favorable technical attributes, the system of school tions. The public companies have also been criticized: buses, minibuses, microbuses, and collective taxis the fixed salary, which does not take into account the offered many more advantages (see Owen 1973; Walters number of trips or passengers, destroys the incentive to 1979). First, a small vehicle allows more intensive use of pick up passengers and to maintain the vehicle in opti- the most abundant factor in developing countries, labor, mum condition. and because the level of investment is lower, vehicles The private system appears to decrease the informa- are easier to obtain, more drivers can own their own tion costs required for efficient operation. That is, since vehicles, and there are greater incentives to maintain the driver has an incentive to pick up as many passen- vehicles in good condition. Second, conditions on the gers as possible, less planning and monitoring are re- demand side make the use of low-capacity vehicles at- quired to ensure that bus schedules and operations are tractive: because the waiting and travel times are shorter meeting passengers' needs. This does not mean that a and there are fewer stops, this type of service is more like fixed salary scheme cannot establish information and that of a private vehicle and is more attractive to middle- control systems for efficient operation. and high-income groups. All these arguments have con- The incidence of the subsidy is often misunderstood. tributed to legitimize the persistence of diverse vehicle The user obtains an economic benefit from the lower sizes. fare or better service even though the subsidy is given to 148 Alvaro Pachon the bus owners. The short-term effect is also sometimes dispersion is the flat fare, not the subsidy, and dispersion confused with the long-term effect. In a system with no can be reduced by making the fare vary according to restrictions on the entry of new buses, an excessive distance traveled. subsidy can lead to large temporary profits for the vehi- cle owners. Free entry ensures, however, that in the long term automotive stock will increase, the service level New versus Old Buses will improve, and profits will become normal. Finally, transfers are often confused with economic costs. A Preferential treatment for more recent models was transport subsidy is a transfer. Even though a subsidy initially provided through differential fares by age of decreases consumer surplus-because resources are vehicle and, more recently, through differential sub- being used at a higher value than what consumers are sidies by model. In October 1980 monthly subsidies were willing to pay-a public transport subsidy tends to cor- 42,812 pesos for 1974-and-later models, 30,443 pesos for rect the inefficient resource allocation caused by the 1970-73 models, 26,687 pesos for 1965-69 models, absence of a price system for use of road space. 25,811 pesos for 1960-64 models, and 24,917 pesos for The absence of empirical research on the economic 1959-and-earlier models. The cFr also favors the financ- effects of government intervention in the transport sec- ing of new vehicles in granting credit. tor is beginning to be remedied. Pach6n (1981b) has found that the effect of a transport subsidy is essentially The Rationale for Differential Treatment redistributional; the benefits reach mainly lower- income households. The maintenance of low fuel prices Differential fares and subsidies by age are defended as has also contributed to a better income distribution by a means of stimulating the demand for new domesti- decreasing the buses' operating costs. A preferential cally produced vehicles. Furthermore, if it is desirable subsidy for bus users has fewer leakages than a general to augment the capital stock rapidly, preferential treat- subsidy for all public transport users. Further research ment for new vehicles is helpful. The operating costs of is needed to quantify the effects of transport subsidies. vehicles of different ages can also be relevant: to the extent that the operating cost of a new vehicle is less Fares and Subsidies than that of an old vehicle, the former will be used more intensively. While the managers press the government for in- There are also reasons for giving equal or similar creases in fares whenever they consider that their invest- treatment to old and new vehicles. Only 7.5 percent of ment yield has reached unsustainably low levels, work- the stock of vehicles in Bogota is less than two years old. ers and students protest when bus fares are increased. In The high proportion of old vehicles with lower capital an effort to lower costs to the public, the govemment costs offers drivers the possibility of owning vehicles and establishes a fare below the cost of service, maintains thus broadens the base of ownership. A balanced treat- low prices for gasoline and diesel fuel, controls the ment for used buses could be defended, since it pro- prices of such inputs as tires, and establishes subsidized motes capital formation, widens to some degree owner- credit lines through the crr. To guarantee an adequate ship of the means of production, and allows people profit, it specifies a subsidy for bus owners and sets whose opportunity cost of time is lower to maintain unsubsidized fares for means of public transport other older vehicles. than buses. Some operational parameters have been defined, and an index of input costs is maintained that allows the Operational Costs, by Age of Vehicle government and the bus owners to recognize when The Interconsult Ltda. study found no statistical dif- periodic adjustments to public transport fares are ference in operating costs between new and old vehicles, needed to reflect increases in costs. Thus, adjustments but since such variables as routes served, distance were made in response to increases in world fuel prices. traveled, and drivers were not isolated, more analysis is There is great confusion, however, regarding the eco- needed. Merewitz (1977) cites two studies which show nomic effects of such measures. Transport subsidies are that the operating costs of a company are independent of sometimes blamed for an exaggerated urban dispersion, the average age of the fleet. As in the Interconsult Ltda. but in a flat fare system, as in Bogota, the marginal cost study, the estimation procedure and the data used have of a kilometer traveled is zero no matter what the fare, limitations. and since decisions on residential location depend on A regression analysis of costs and revenue in relation the marginal cost of transport, a transit subsidy has no to age has been carried out with the use of assumptions effect on housing location. Thus, the cause of urban for the estimation of operating costs by ranges of bus Urban Transport Policy: Colombia 149 models and data on the monthly operating costs for clusions. First, variations in the operating costs per different bus models as estimated for October 1980.' The kilometer of buses of different ages are relatively small. following relations were calculated with the use of Second, because the monthly operating revenue by age ordinary least squares. decreases faster than the monthly variable costs by age, (12-1) In Y, = In Y0 + Ba older buses tend to be less competitive. Third, the num- ber of monthly trips decreases with age-new buses where Y is the variable whose behavior is to be analyzed, seem to be used more intensively. Fourth, additional a is the vehicle's age, and In Y0 and B are the function's research is needed on the cost and income structure by parameters. age of the vehicle, with the effect of such other variables The number of trips made, which varies inversely with as route conditions, driver characteristics, and the make age, influences the number of kilometers traveled and of the vehicle isolated. the monthly cost. To isolate the effect of trips made, the variable cost per kilometer (VCK) was calculated and a Prices and Profitability regression was run, with the following results.2 (12-2) In VCK, = 27.925 + 0.0024 + a . For used vehicles the present value of future income is (4.807) basic to the price determination and in turn to evalua- R2 = 0.9001 tion of the effects of fare and subsidy policies. The price pattern of buses by age is satisfactorily approximated by Although the VCK changes very little (it decreases 1 an exponential functional form for cars in Colombia percent for every fouryears of age), the hypothesis that it (Pach6n 1981a). Data on the value of the vehicles have is independent of age cannot be accepted because the recently been assembled by the Center for Economic t-statistic value is greater than the critical value. Development Studies (Centro de Estudios para Desar- The results of monthly cost and revenue regressions rollo Econ6mico, CEDE) in Colombia, and these data are were as follows, where VC is variable cost, FC is fixed used to estimate the following form. cost, and ING is operating revenue. (The subsidy is not counted as revenue.) (12-6) In Pa, = In PO - b *a (12-3) In VCa= 11.3088 - 0.01806 +a where Pa is the price of a vehicle of agea. The results of (8.37) the regression are R = 0.9589 (12-7) In Pa = 14.3961 - 0.06887a (12-4) In FCa = 8.8097 + 0.02771 + a (141.1) (8.28) (14.13) R2 = 0.7742 = 0.9852 20 degrees of freedom (12-5) In INGa = 11.2157 - 0.02167 + a The t-values are in parentheses. (6.65) = 0.9365 The interpretation of the age variable coefficient in a R = 0.9365 regression of the price logarithm against the age of the The results show that both monthly cost and revenue are vehicle is the yearly depreciation rate. According to the negatively affected by age and that the coefficient of the results given here, for each year of use the price of age variable is statistically different from zero. Accord- the bus decreases by slightly less than 7 percent. The ing to these results, monthly income decreases more value obtained in equation 12-7 is less than that used in rapidly than monthly variable costs, and the intercept of the cost studies made by the transit companies, which the variable cost function is slightly greater than the show a useful life of seven years and a salvage value of 30 corresponding value of the income functions. Thus, percent.3 Thus the depreciation rates chosen in the stud- operational revenue does not cover the variable costs, ies on defining the fares tend to produce real profit rates and the gap between revenue and costs increases in higher than those agreed on between the government percentage terms as the vehicle gets older because in- and the bus owners. come decreases faster than costs. The economic profitability of used vehicles is affected Table 12-1 presents results for each model between by three sources of economic rent: depreciation (the loss 1959 and 1980. The adjusted values were calculated in value owing to age), the opportunity cost (the mone- using equations 12-3, 12-4, and 12-5. To obtain capital tary interest rate multiplied by the price of the good), return, the estimated income and the subsidy received and the appreciation of the good because of inflation and were added and total costs were subtracted. changes in relative prices. Given estimates, by age, of The evidence on operating costs suggests several con- depreciation and the price of vehicles, a nominal profit- 150 Alvaro Pach6n Table 12-1. Buses: Adjusted Costs, Estimated Incorne, and Capital Return, October 1980 (pesos) Estimated Adjusted variable variable Total Estimated Subsi- Capital Model Age costsa costsb costs income' dies return 1959 21 55.800 3.743 59.543 47.126 24.917 12.400 1960 20 56.817 3.848 60.665 48.158 25.811 13.303 1961 19 57,852 3.957 61.809 49.213 25.811 13.215 1962 18 58.907 4.068 62.975 50.292 25.811 13.127 1963 17 59.980 4.182 64.163 51.393 25.811 13.041 1964 16 61.074 4.300 65.374 52.519 25.811 12.956 1965 15 62.187 4.420 66.008 53.670 26.687 13.749 1966 14 63.320 4.545 67.855 54.845 26.687 13.667 1967 13 64.474 4.672 69.147 56.047 26.687 13.586 1968 12 65.649 4.804 70.454 57.275 26.687 13.508 1969 11 66.846 4.939 71.785 58.530 26.687 13.431 1970 10 68.064 5.077 73.142 59.812 30.443 17.112 1971 9 69.305 5.220 74.525 61.122 30.443 17.039 1972 8 70.568 5.367 75.935 62.461 30.443 16.968 1973 7 71.855 5.517 77.372 63.829 30.443 16.900 1974 6 73.164 5.672 78.837 65.228 42.812 29.202 1975 5 74.498 5.832 80.330 66.657 42.812 29.138 1976 4 75.856 5.996 81.852 68.117 42.812 29.077 1977 3 77.238 6.164 83.403 69.609 42.812 29.018 1978 2 78.646 6.337 84.984 71.134 42.812 28.962 1979 1 80.079 6.515 86.595 72.693 42.812 28.909 1980 0 81.539 6.698 88.238 74.285 42.812 28.859 a. VCa = 81,539 [exp(- 0.0180a)] R2 = 0.9589 b. Fir, = 6,698 [exp (- 0.0216a)l R2 = 0.9852 c. INGa = 74.285 lexp(- 0.0216a)] R2 = 0.9365 ability rate can be established once the rate of increase in sarily high. As this occurs for the older buses, the ad- bus prices is known. Estimated prices, capital return justed prices underestimate the prices of the older vehi- (current revenue minus current cost), and profitability cles and consequently overstate their profitability. yielded by the calculation are presented in table 12-2. To show that many of the distortions in prices and The estimated values are based on regressions in which profitability are a result of the subsidies, data for buse- the only explanatory variable is the vehicle's age and tas, which are not subsidized, were used. With the use of which therefore may differ from real values. the buseta price data from the CEDE survey, an exponen- The estimated profitability (ia) shown in table 12-2 is tial function was estimated: inversely related to age. This finding is quite unex- (12-8) In Pa = 14.0933 - 0.06943*a pected. Under competitive conditions, profitability (141.1) (8 .63) might be considered to be independent of age. A casual ( ( R2 - 0.8515 observer, however, might expect higher profitability for 14 degrees of freedom the new buses, since larger owners tend to own newer buses. Two possible explanations may be offered for the In table 12-3 the buseta's profitability is nearly the same results in table 12-2. First, the difference could be a for 1970-and-later models; for previous models profit- result of an underestimation of the operating costs of old ability decreases with age. Thus, in the absence of a models. The maintenance costs of old vehicles exclude subsidy the estimated profitability is more in line with the value of the time of the driver, who often owns and an expectation that profitability is independent of age. repairs the bus. Second, the difference could be a con- sequence of an underestimation of the price of older vehicles. Because the estimated prices are calculated on Buses or Busetas? the basis of a regression whose only explanatory variable is the vehicle's age, these prices cannot capture the Govemment policy appears to have favored busetas, as distortions caused by the subsidy and thereby tend to is reflected in their rapid growth. Between 1971 and underestimate cases in which the subsidy is unneces- 1976 the number of buses is estimated to have grown by Urban Transport Policy: Colombia 151 Table 12-2. Buses: Prices, Capital Return, and Profitability (pesos) Price Estimated Capital Profit- Model Age in 1980 pricea retumb ability' 1959 21 200,000 420,759 12,499 44.15 1960 20 540,000 450,758 13,303 43.92 1961 19 640,000 482,896 13,215 41.34 1962 18 656,000 517,325 13,127 38.96 1963 17 666,000 554,209 13,041 36.74 1964 16 600,000 593,723 12,956 34.69 1965 15 635,000 636,053 13,749 34.44 1966 14 109,000 681,402 13,667 32.57 1967 13 650,000 729,985 13,586 30.84 1968 12 719,000 782,031 13,508 29.23 1969 11 650,000 837,788 13,431 27.74 1970 10 894,000 897,520 17,112 31.38 1971 9 893,000 961,511 17,039 29.77 1972 8 733,000 1,030,064 16,968 28.27 1973 7 1,100,000 1,103,505 16,900 26.88 1974 6 1,400,000 1,182,182 29,202 38.15 1975 5 1,342,000 1,266,469 29,138 36.11 1976 4 1,382,000 1,356,765 29,077 34.22 1977 3 1,480,000 1,453,499 29,018 33.33 1978 2 1,475,000 1,557,130 28,962 30.82 1979 1 1,617,000 1,668,149 28,909 29.30 1980 0 1,933,000 1,787,084 28,859 27.88 a. PO = 1,787,084 [exp (0.0688a)1 In P, = In P0 14.391 - 0.06887a (141.10) (8.28) R2 = 0.7742 where depreciation (a) = 6.9 percent. b. Capital return (RK) = income - costs. c. AssumingthatRKa = (a + i± - ar)Pa, thenia = (RRKaPa) - a + Tr,wherei4isprofitability, a = 6.9 percent (seeequation 12-8), and r = 15.41 percent. Source: 1980 prices: Jaramillo (1981). 2.6 percent annually; for busetas in urban areas the rate with the preference for these vehicles on the demand was 30 percent. Between 1971 and 1980 the average side, explains their rapid growth in Colombia. growth rates were 6.2 percent for the bus stock in Bo- The difference in the profitability of each type of gota and 25 percent for the buseta stock; in Cali the vehicle not only is consistent with the rapid growth of growth rates were 2.4 percent for buses and 21.93 per- the buseta stock but also reflects restrictions on the cent for busetas. addition of busetas to the stock. If there were no restric- Walters (1979) has indicated that in developing coun- tions, the profitability of both types of vehicles could be tries minibuses are more efficient than buses with re- expected to be equalized. As mentioned above, a license, spect to costs, waiting time, and trip frequency. Pach6n or nota opci6n, is required before a vehicle can be (1981c) concludes that the demand for busetas increases affiliated to a transport company. To the extent that with income. Furthermore the income elasticities of obtaining this license is more difficult for busetas than demand for busetas based on time series are higher than for buses, a higher profitability for busetas is ensured. those obtained from cross-sections (see Pach6n 1981b). Data on the prices of used vehicles confirm the exis- During the period considered, the number of micro- tence of greater restrictions on the entry of busetas. buses, which are smaller than busetas, decreased. Automotive vehicles show rapid depreciation in the first year and a constant rate of depreciation for the remain- Profitability of Buses and Busetas ing useful years (see Wykoff 1970, pp. 171-72). With such a pattern the adjusted value based on the regres- For 1967-and-later models, the busetas' profitability sion would be less than the value charged by the distrib- is higher than that of buses of the same age.4 This implies utor, since the owner of the vehicle has to pay a penalty that policies regarding public transport fares and sub- to take the vehicle from the distributor. When, on the sidies favor intermediate-size vehicles. This, together contrary, there are supply restrictions it is to be expected 152 Alvaro Pach6n Table 12-3. Busetas: Capital Return and Profitability (pesos) Price Estimated Capital Profit- Model Age in 1980 pricea retumb ability' 1959 21 n.a. 307,202 3,588 23.65 1960 20 n.a. 329,289 4,326 25.40 1961 19 n.a. 352,965 5,098 26.97 1962 18 n.a. 378,343 5.906 28.37 1963 17 n.a. 405,545 6,752 29.61 1964 16 n.a. 434,703 7,636 30.72 1965 15 566,000 465,958 8,560 31.68 1966 14 624,000 499,459 9,526 32.52 1967 13 475,000 535,370 10,535 33.25 1968 12 550,000 573,862 11,580 33.87 1969 11 495,000 615,122 12,688 34.39 1970 10 678,000 659,349 13,836 34.82 1971 9 605,000 706,755 15,034 35.16 1972 8 789,000 757,570 16,283 35.43 1973 7 650,000 812,038 17,585 35.62 1974 6 900,000 870,423 18,944 35.75 1975 5 965,000 933,005 20,359 35.82 1976 4 994,000 1,000,087 21,834 35.83 1977 3 n.a. 1,071,992 23,370 35.80 1978 2 1,375,000 1,149,067 24,971 35.71 1979 1 1,321,000 1,231,684 26,637 35.59 1980 0 1,270,000 1,320,240 28,372 35.42 n.a. Not available. a. Pa = 1,320,240 lexp( -0.0694a)] In Pa = 14.0933 - 0.06943a (141.10) (8.63) R2 = 0.8515 where depreciation (a) = 6.9 percent. b. Capital return (RK) = income - costs. c. Assuming thatRK, = (a + ia - ir)P0, then ia = (RKa/Pa) - a + 7r, where i4 is profitability, a 6.9 percent (see equation 12-8), and iT = 16.54 percent. Source: 1980 prices: Jaramillo (1981). that the owner will obtain a bonus for taking the vehicle If the rapid growth of the buseta stock is seen in from the distributor. In the first case the adjusted price perspective, it can be inferred that in the beginning (using the regression) will be less than the list price; in there were more restrictions on the entry of busetas and the second case the adjusted price will be greater than that the gap between the profit rates for buses and the list price. If the relation between the estimated price busetas was larger. Kozel (1981) has shown that supply and the list price is defined as a premium factor, the restrictions decreased during 1972-78. Essentially, degree of restriction in the market can be inferred be- busetas have been allowed to operate on the same routes cause the bigger is the premium factor, the higher is the as buses, and a greater adjustment of the buseta supply degree of restriction in the market. has been facilitated. For Bogota in 1980 the bus premium factor, Bb, is 1.02; the premium factor for the buseta, Bbt, is 1.28. The Load Factor Both vehicles receive a bonus for leaving the distributor. This factor, for both buses and busetas, tells us that the From the point of view of economic efficiency, where government, through INTRA, still imposes severe restric- there are different types of transport vehicles the equip- tions, in relation to the demand for public transport, on ment should be assigned in order of efficiency, with the the entry of these vehicles to the stock. In the case of most efficient assigned first. The most efficient vehicles busetas the restriction is even clearer, which is in would be expected to work all day and the least efficient accordance with their high profitability, and the supply only during peak hours. Where there are two types of of this mode should be even higher to obtain equilib- equipment with different fares, assignment would rium between supply and demand. generally be in order of increasing break-even load fac- Urban Transport Policy: Colombia 153 tors: the equipment with the lowest break-even load Observedloadfactor Break-even loadfactor factor would be assigned first and that with the highest Vehicle (percent) (percent) break-even load factor last. Such an operating rule Bus 69 78.5 causes the vehicle with the lowest break-even load factor Buseta 66 43.2 also to have the lowest observed load factor. The results obtained in this section show that the fare In Colombia, since the bus owner receives a monthly and subsidy policies encourage a more intensive use of subsidy and the buseta owner receives no subsidy, the busetas, which favors the buseta supply in both the short buseta's break-even load factor should be lower.5 In and the long run. consequence the buseta would tend to be used during both peak and nonpeak hours and to have a lower load factor. Summary and Conclusions With the use of assumptions for calculating the bus and buseta load factors in Bogota, and cost data for some In Colombia the state has tried to favor public over periods when fares were changed, the effect of these private transport. Given the absence of policies-other changes on the break-even load factors can be seen. For than the subsidy for buses-to remedy the externalities example, the bus break-even load factor increased from of private transport through the price system, it may be April 1978 to May 1980, which reflected the slower necessary to increase the use of quantitative restrictions. increase in fares than in variable costs. After the most For example, some lanes may be dedicated to the exclu- recent fare increase studied, in October 1980, the break- sive use of public transport. Simultaneously with these even load factor decreased, since fares had increased quantitative restrictions, other policies that use the more than costs. It can be inferred that the use of buses price system to discourage the use of private vehicles had been decreasing between April 1978 and October should be investigated. The proposals for a fuel tax and a 1980, perhaps because the use of busetas increased. parking tax to reduce congestion should be studied The results show the dilemma between efficiency and further. In addition, the possibility of establishing a equity which the government confronts in fixing fares. If license for access to the central zones, as in Singapore, the government wants to decrease the impact of the should be examined. growth of operating costs by moderating the increase in The combined effect of the bus subsidy, the implicit fares and increasing the subsidy, the result may be fuel subsidy, and the vehicle tax has been to favor the stimulation of the use of vehicles that occupy more road lower-income classes. But since the private companies space and that consume more fuel per passenger trans- have much lower operating costs than does the district ported. government company, it is probable that increases in It can also be seen that although the load factor for the allocational efficiency attained from a system with all buseta increased between May and October 1980, there government vehicles would increase operating costs. is a great difference between bus and buseta load factors. The route assignment process is flexible and has been Whereas a bus must fill 78.5 percent of its capacity able to adapt itself to changes in the location of activi- to cover its operating costs, the buseta needs to fill only ties. 43.2 percent of its capacity. This great difference in load Such problems of public transport as slow operating factors explains why a bus owner prefers to wait for peak speed can be eliminated by such measures as estab- hours to operate his bus. Only during those hours is lishing posted bus stops on all lines, reserving some there enough demand to guarantee that operating costs roads for exclusive use by public transport, and charg- will be covered and a profit made. The buseta, in con- ing, by tender, for the use of roads dedicated exclusively trast, can operate on routes with low demand and during to public transport. nonpeak hours because it needs to use only 43.2 percent The credit established by the cFr has made investment of its capacity to cover its costs. This low factor suggests in public service vehicles easier. The availability of that the buseta's trip frequency will be greater than that credit, together with the fare and subsidy scheme, has of the bus because it does not have to wait for peak hours assured an adequate profit rate for investment in differ- to operate profitably. ent types of vehicles. The driver's remuneration system If we compare the observed load factor with the re- creates incentives for the efficient operation of the vehi- quired load factor as in the table below, we find that the cles but can also contribute to traffic control problems. load factor for the bus is too low to cover operating costs, As has been proved in Cali, the authorities can make while that of the buseta is distinctly above the break- substantial improvements through campaigns directed even load factor. This result tells us that the buses are at users. incurring losses or that their operating costs have been The elaborate system for setting fares and subsidies overestimated. has been described above. The adoption of a methodol- 154 Alvaro Pach6n ogy to calculate a transport cost index and the periodic reduce the distortions from information problems or collection of the required data have permitted relatively methodological limitations. For example, by eliminat- frequent adjustments that have maintained adequate ing restrictions on the entry of public transport vehicles, profits for the investment in public transport vehicles the same profitability would be ensured for all types of despite inflation and changes in fuel prices. The flat fare vehicles, and transport would be as profitable as other scheme was identified as a possible cause of excessive sectors of the economy. Similarly, some fares, such as urban dispersion. It was shown that the subsidy policies night service charges, could be freed from controls. unnecessarily increase the price of old vehicles and arti- Considerable knowledge exists concerning urban ficially raise the prices of models at the lower end of the transport in Colombia, and the use of models to simulate range to which the subsidy applies. It was further shown routes is fairly well established. Government transport that fare and subsidy policies favor investment in buse- policies have been, in general, in the right direction. tas and a more intensive use of these vehicles. The Policies have sought to give priority to public transport restrictions on the entry of busetas into the stock in and have, by and large, favored lower-income classes. 1980 were greater than was the case for buses, and it was noted that this fact is compatible with a decrease in restrictions for buseta entry during 1972-78. The results of the research on the economic effects of Notes the fare and subsidy policies lead to the conclusion that the present scheme is causing distortions in the supply 1. The information on costs may be overestimated because of transport vehicles. Some simple reforms of the sub- of the nature of the data and their relation to fares. The sidy scheme could eliminate the distortions and at the FEDESARROLLO (Urrutia 1981) study used the operating costs of same time solve other problems. A first proposal is to its buses and obtained lower operating costs. change the basis for the subsidy from model to age, to 2. The t-values are in parentheses, there are three degrees of freedom, and the critical t-value for three degrees of freedom is avoid the model effect and the price distortion of old 3.182. vehicles. The same profit rate would be adopted for buses 3. The t-value to prove the hypothesis b = 0.1 is 2.846. It is and busetas. As in the present system, only buses would greater than the critical value 2.086 with a 95 percent confi- be subsidized. dence level and 20 degrees of freedom. In this analysis the values for buses and busetas were 4. The profitability of buses can be less than that observed to calculated. In both cases a 40 percent annual profit rate the extent that there are delays in the payment of the subsidy by in monetary terms was fixed. The bus fare covers operat- the cFT, the entity in charge of the payment. ing costs, and the subsidy covers capital and fixed costs. 5. The buseta owner must cover both variable and fixed Given a 40 percent profit rate, a uniform fare for a bus of costs from fares, while the bus owner covers his costs out of the any age is 3.50 pesos. The present subsidy is unneces- subsidy and the fares. The buseta owner therefore needs a sarily high for older buses. It is possible to fix a flat fare of lower load factor to cover his variable costs. 7.80 pesos for busetas, a little higher than the present fare (7.50 pesos), to ensure a nominal profit rate of 40 Bibliography percent, equal to that of buses. A second reform would allow additional benefits. Cifuentes, Jorge Ignacio. 1978. "Urban Transportation in There is a desire in Colombia to maintain the price of Bogota." Water Supply and Urban Development Depart- fuel used by public transport below that of fuel used by ment, World Bank, Washington, D.C. Processed. private transport. The authorities have wanted to change Corporaci6n Financiera del Transporte. 1979. Aspectos esta- all the public stock to diesel and to maintain low prices disticos del parque y las empresas de servicios de transporte for that fuel, but this policy has not had much success. A colectivo urbano. Bogota. possible measure is a subsidy that is related to the Departamento Nacional de Planeaci6n-Instituto Nacional de average fuel consumption for each type of vehicle used Transporte (INTRA). 1974. Estudio de organizacidn y admin- in public transport. If the proposal to adopt a subsidy istracidn deltransporte colectivo urbano deBogotd (oAT II). scheme that would eliminate some of the distortions Bogota. created by the present scheme is accepted, additional INTRA. 1980. Informe final sobre el programa de indices de research on the vehicle's costs, income, and price by age precios de los insumos del transporte. Bogota. will be needed. A better level of information would per- Interconsult Ltda. 1970. "Estudio de organizaci6n administra- mit subsidies to be based on the costs of different ages of tiva del transporte colectivo de Bogota" (ROT). Bogota. vehicles without incurring serious distortions. Jaramillo, Samuel. 1981. "Situaci6n de los conductores del Along with improvements in the definition of fares transporte colectivo urbano en Bogota." Estudios laborales and subsidies, market forces should be permitted to no. 1. Desarrollo y sociedad. Bogota. Urban Transport Policy: Colombia 155 Kozel, Valerie. 1981. "Travel Demand Models for Developing .1981b. El impacto redistributivo de la intervenci6n del Countries: The Case of Bogota, Colombia." Urban and Re- gobiemo en el transporte." Documento de trabajo 19. Bogo- gional Report 81-26. World Bank, Washington, D.C. Pro- ti: Corporaci6n Centro Regional de Poblaci6n. cessed. . 1981c. "Situaci6n del transporte en Cali y Bogota, Leibenstein, Harvey. 1966. "Allocative Efficiency vs. X decada de los setenta." Documento de trabajo 18. Bogota: Efficiency." American Economic Review, vol. 56 (June), Corporaci6n Centro Regional de Poblaci6n. pp. 392-415. Republic of Colombia. 1973. Fiscal Studies. Bogota Urban McClure, Charles. 1974. "Incidencia de la tributaci6n en Co- Development Study, Phase 11. Bogota. lombia." In Gilis Musgrave, ed., Propuesta de una re- Urrutia, Miguel. 1981. "Evaluaci6n del sistema de transporte forma fiscal para Colombia, vol. 2, Bogota: Banco de la publico en Bogota." In Buses y Busetas-Una evaluacion Republica. del transporte urbano de Bogotd. Bogota: FEDESARROLLO. Merewitz, Leonard. 1977. "On Measuring the Efficiency of Walters, A. A. 1979. Costs and Scale of Bus Services. World Public Enterprises: Bus Operating Companies in the San Bank Staff Working Paper 325. Washington, D.C. Francisco Bay Area." Transportation (March), pp. 45-55. Walters, A. A., and Charles Feibel. 1980. Ownership and Nelson, C. 1972. "An Econometric Model of Urban Bus Transit Efficiency of Urban Buses. World Bank Staff Working Paper Operations." In J. D. Wells and others, ed., Economic Char- 371. Washington, D.C. acteristics of the Urban Public Transportation Industry. Westin, B. Richard. 1980. "Travel Pattems in Bogota: A De- Washington, D.C.: Institute for Defense Analyses, for the scriptive Analysis for 1972." Water Supply and Urban De- U.S. Department of Transportation. Available as PB 265-325 velopment Department, World Bank, Washington, D.C. Pro- from National Technical Information Service. cessed. Owen, Wilfred. 1973. Automobiles and Cities: Strategies for World Bank. 1975. "Transportes urbanos." Documento de Developing Countries. World Bank StaffWorking Paper 162. politica sectoriaL Washington, D.C. Washington, D.C. Wykoff, Frank C. 1970. "Capital Depreciation in the Postwar Pach6n, Alvaro. 1981a. "El autom6vil en dos metr6polis del Period: Automobiles." Review of Economics and Statistics, tercer mundo." Documento de trabajo 17. Bogota: Corpora- vol. 52 no. 2 (May), pp. 168-72. ci6n Centro Regional de Poblaci6n. 13 Evaluating Pollution Control: The Case of Sdo Paulo Vinod Thomas S ao Paulo is the most industrialized area of Latin The Extent and Sources of Pollution SAmerica, and after Mexico City and Shanghai it is Pollution in Sao Paulo has been studied by, among the third largest metropolitan center in the world.' To- othes Ee ado Pannesc (18) s ianneschi, day some of the most urgent problems of the urban others, Esteves and G Lanneschi (1980); Gianneschi, environment are found here, particularly in the highly Junior, and Salvador g1979); Licco, Oda, and Galvao industrialized sections of the Greater Sao Paulo Metro- World Bank. politan Area (GSP). Air pollutant discharges in the GsP currently amount to 8,000 tons daily. Carbon monoxide constitutes 65 Pollution: Problems and Policies percent of the emissions, sulfur dioxides 13 percent, hydrocarbons 10 percent, particulates 7 percent, and Until the early 1970s Brazilian policymakers tended nitrogen oxides 5 percent. In 1978 daily air quality to opt for environmental damage in the pursuit of eco- standards were exceeded 299 times for carbon monox- nomic growth and competitiveness in international ide, 121 times for particulate matter, and 17 times for markets. As pollution grew, however, the virtues of sulfur dioxide. The highest concentrations of these sub- accepting such a tradeoff-even if it existed-became stances in the same year were, respectively, 282 percent, questionable. Under growing pressure from localities 190 percent, and 219 percent more than their standards. national antipollution policies, supported on the state During 1976-78 the average annual concentration of and metropolitan levels, began to evolve in the mid- carbon monoxide in the more polluted areas was 12- 1970s. The enactment of laws is of more recent origin, 13.5 parts per million (ppm), while that of particulates and actual measures have been mostly limited to the GSP. was 115-126 micrograms per cubic meter. These levels As policies are now beginning to be implemented, how- are comparable with those recorded in some of the ever, old anxieties have re-emerged about their possible largest cities in the United States during the early effects on competitiveness in world markets, on infla- 1970s.2 tion, on efficient use of energy, and on the government The damage is compounded by Sao Paulo's climate budget. The shifts in opinion about environmental reg- and topography. Light winds, air stagnation, and fre- ulations call attention to the need for objective analysis quent temperature inversions hold polluted air close to and estimates of the effects of policies. This chapter uses the ground. In the winter months of June, July, and a cost-benefit framework to analyze the welfare effects of August the concentration at ground level rises as much industrial location and pollution control, with the focus as 50-75 percent over the annual mean in some places, on industrial air emissions. which aggravates respiratory illnesses. Note: This chapter has been published in a somewhat different form Spatial variations in damages are sharp within the as "Evaluating Pollution Control," Joumal of Deve(opment Eco- thirty-seven municipalities in the GSP. Nine municipali- nomics, vol. 19 (1985), pp. 13346. ties contribute about 95 percent of total air emissions. 156 - Evaluating Pollution Control: The Case of Sao Paulo 157 An analysis of the lead content in the blood of humans nipulating the growth of industrial output. This policy compared three neighboring areas: Embu-Gaucu, which has resulted partly from the current strategy of trying to is mildly polluted; Sao Paulo municipality, which is prevent a worsening of the problem rather than reduce more polluted; and Sao Bernardo, which is heavily pol- existing emissions. Toward this goal, the locational deci- luted (Fernicola and Azevedo 1979). The average lead sions of new producers have been influenced through concentrations in blood among representative groups of licensing and zoning procedures. The CETESB now people are shown below. licenses all new industrial investment in the state of Sao Paulo. Average lead In the future, options other than those that directly in blood (micrograms Standard affect output may be expected to increase. Already there per cubic meter) deviation is a $187 million project for air and water pollution Embu-Guacu 11.2 5.6 abatement which regulates a small number of industrial Sao Paulo municipality 12.4 4.8 plants that account for a very high percentage of total Sao Bernardo 20.5 5.7 pollution. The air pollution control is aimed primarily at In the csp, about 94 percent of carbon monoxide, 73 particulates, and the water pollution control focuses on percent of nitrogen oxides, and 72 percent of hydrocar- toxic wastes. Loans are offered to enable plants to meet bons are discharged by some 2 million vehicles. Indus- state and federal quality standards. It is chiefly plants in trial processes and stationary fuel combustion by ap- the GSP that qualify for the loans. In operation the pro- proximately 30,000 establishments are responsible for ject is expected to stress whatever tums out to be the about 80 percent of particulate matter, over 90 percent least-cost control strategy for reducing smoke and meet- of sulfur dioxide emissions, 19 percent of hydrocarbons, ing air quality standards. Since air quality is measured in and 25 percent of nitrogen oxides. Automotive traffic is the area near a factory (rather than in the smokestack), clearly the largest source of total pollution, while in- greater efforts and higher control costs are to be ex- dustrial processes and stationary fuel combustion are pected in the more polluted areas. the principal sources of particulates and sulfur dioxide. Roughly 65 percent of the industrial emissions are re- leased by a relatively few industries in the nonmetallic, The Costs of Pollution Control chemical, and metallurgic categories. According to Gianneschi, Junior, and Salvador (1979), about 90 per- Estimates of the capital and operating costs of pollu- cent of particulates and 74 percent of sulfur dioxide are tion control equipment and spare parts for 285 indus- discharged, respectively, by only 5 percent and 10 per- tries that contribute 97 percent of industrial particulate cent of the sources; at the other extreme, 3 percent of pollution in the GSP have been compiled. (These esti- the particulates and 10 percent of the sulfur dioxide mates are CETESB projections that used information for come from 75 percent and 73 percent, respectively, of industries without control equipment.) For the fifty- the least polluting industries. three industries that account for 90 percent of particu- late effluents, a 94 percent abatement-that is, 85 per- The Policy Focus cent of industrial particulates and 55 percent of all particulates-might be achieved in 1977 at a total cost of Federal legislation has established uniform national about 418 million cruzeiros (about $30 million) or about air quality standards and broad guidelines for their 3,200 cruzeiros ($225) a ton of emissions.3 If a ten-year achievement (Companhia Estadual de Technologia de life of the equipment and a 10 percent interest rate are Saneamento Ambiental [CETESBI 1979a, b; Empresa Met- assumed, the annualized investment cost is $4.75 mil- ropolitana de Planejamento de Sao Paulo [EMPIASA] lion. Addition of a (liberal) 25 percent for annual labor 1979; Pazzaglini and Greco 1981). By and large, uniform and maintenance costs yields an overall annual cost of abatement by polluting industries has also been sought about $6 million. If 25-50 percent of the GSP'S 11 million but thus far has been enforced unevenly in different people are assumed to be directly affected by pollution,4 a areas. Enforcement of abatement has been effected per capita annual cost of about $1.1-$2.2 is implied. through the CETESB, the state's pollution control agency, Elimination of an additional 7 percent of smoke from with a system of fines and penalties which has been more 232 more industries (or, the collection of about 91 effective in the csP than elsewhere. Thus, at the moment percent of industrial particulates and about 70 percent there is considerable spatial differentiation in pollution of all particulate effluents) more than doubles the total abatement. cost, to about 964 million cruzeiros ($68 million in One approach adopted to contain pollution has been 1977), and the cost a ton to 6,800 cruzeiros ($480). to indirectly influence environmental quality by ma- The Medio Paraiba area, which has a population of 158 Vinod Thomas over 1 million, is a small but heavily industrialized and declines somewhat with the volume of production, X, as heavily polluted area in the state of Sao Paulo.s Approxi- shown in the following examples. mate total particulate emissions from point sources are e, in kilograms over 42,466 tons a year. According to Kowalczyk (1980) Industry per ton of output X the cost of the best control technology, which achieves Steel: basic oxygen process 6X about 98 percent abatement, is $16.6 million (in 1979 Iron 9X dollars), or about $550 a ton if the technology is applied Cement: calcination fumace 6X to the four sources which account for 72 percent (30,759 Ceramics 9.5X58 tons) of all industrial pollution in the area. If 25 percent Ammonium phosphate l0x,8 annual labor costs are assumed, an annualized invest- If the emission factor declines significantly with output, ment plus labor cost figure is $3.25 million to abate it would be advantageous to require less than propor- 30,759 tons annually. If only 75 percent of the people are tionate abatement from larger producers and thereby directly injured by the airborne discharges, this means allow those who pollute less per unit of output to pro- an annual cost of $4.33 (1979 dollars) for each person duce more. affected. Economies of scale in pollution control are implied in Licco, Oda, and Galvao Filho (1979) estimate that in Kowalczyk (1980). As table 13-1 shows, the control costs many Brazilian industries effluent per unit of output per ton of emissions for the largest polluters, who con- (the emission factor e) is constant. In other industries it tribute 90 percent of the particulates in the GSP, are Table 13-1. Control Costs for Industries That Contribute 97 Percent of Particulate Emissions in the GSP (thousands of cruzeiros, except where otherwise specified) Control costs (1977 estimates) Number Emissions Control Capital costs Installation costs Total cost Total cost of (tons efficiency Per Per Per (1980-81 Industry industries' per year) (percent) Total ton Total ton Total ton estimate) Nonmetallic 110 107,015 n.a. n.a. n.a. n.a. n.a. 443,870 4.15 1,194,600 Average 973 n.a. n.a. n.a. n.a. n.a. 4,035 4.15 10,860 Top 90 percent 36 103,770 95 110,810 1.07 101,620 0.98 212,430 2.05 660,340 Average 2,882 95 3,078 1.07 2,822 0.98 5,900 2.05 18,340 Highb 9,600 n.a. 10,000 1.04 10,000 1.04 20,000 2.08 54,000 Lowb 400 n.a. 2,600 6.50 1,600 4.00 4,200 10.50 11,300 Chemicals 20 16,030 n.a. n.a. n.a. n.a. n.a. 122,190 7.62 328,900 Average 802 n.a. n.a. n.a. n.a. n.a. 6,110 7.62 16,455 Top 90 percent 6 15,360 80 n.a. 2.81 31,900 2.08 75,100 4.90 202,000 Average 2,560 80 7,200 2.81 5,317 2.08 12,517 4.90 33,670 Highb 3,300 n.a. 6,550 1.98 5,200 1.59 11,800 3.57 31,760 Lowb 980 n.a. 6,550 6.68 3,950 4.03 10,500 10.71 28,300 Metallurgic 115 17,110 n.a. n.a. n.a. n.a. n.a. 397,270 23.22 1,069,170 Average 110 n.a. n.a. n.a. n.a. n.a. 2,563 23.22 6,898 Top 90 percent 11 12,560 98 68,000 5.41 62,000 4.94 130,000 10.35 350,110 Average 1,142 98 6,182 5.41 5,636 4.94 11,818 10.35 31,800 Highb 2,200 n.a. 10,700 4.86 10,700 4.86 21,400 9.72 57,590 Lowb 430 n.a. 3,000 6.98 2,000 4.65 5,000 11.63 13,460 All industries 285 140,155 n.a. n.a. n.a. n.a. n.a. 964,330 6.88 2,592,670 Average 492 n.a. n.a. n.a. n.a. n.a. 3,384 6.88 9,097 Top 90 percent 53 101,690 98 220,401 1.68 195,520 1.48 417,530 3.16 1,212,450 Average 3,485 94 4,190 1.68 3,690 1.48 7,878 3.16 22,800 n.a. Not available. a. Particulates involve a total of about 1,150 enterprises. b. Excluding extremes. Source: CETESB estimates; Kowalczyk (1980). Evaluating Pollution Control: The Case of Sdo Paulo 159 significantly lower than for the other polluters. Thus, it Thus, if the national air quality standard of 9 ppm for may be advantageous to focus control policies on the carbon monoxide were met, COHb would be kept to largest polluters. If an 85 percent overall abatement about 2 percent for nonsmokers. The relation between were achieved, the effort to control the remaining 10 increasing cardiovascular disease and COHb content of percent-the smaller polluters-would raise overall over 2 percent has been cited in Femicola and Lima costs, compared with a strategy of relying only on the (1979). top 90 percent of the polluters (assuming a control Mendes (1976) has examined the effect of air pollution efficiency of 94 percent).6 on mortality in an epidemiological study that correlated The efficiency of control equipment and the abate- deaths for each day in the GSP in 1973 with meteorologi- ment cost for each unit also vary among industries (see cal conditions and air pollution levels. Five peaks in table 13-1). Nonmetallic industries, which contribute deaths were observed; at least one, on August 1, was the bulk of particulate emissions (76 percent excluding clearly related to a dramatic worsening of air pollution. fuel combustion), face significantly lower control costs Particulate and sulfur dioxide concentrations rose after than do chemical and metallurgic industries. (Industries July 25 and reached very high levels by July 30. On excluded from the table experience still higher control August 1 sulfur dioxide reached a peak in Capuava of 452 costs.) micrograms per cubic meter, in Aclimacao of 371, in The range in cost differences is apparent when a low Tatauape of 292, and in Cerqueira Cesar of 288. On the control cost industry, cement, is compared with iron, a same day deaths reached a high for the year of 299, high control cost industry (see table 13-2). As a greater compared with an annual average of 228. Deaths of variety of equipment becomes available in the market people over 65 or less than 1 year old and deaths attrib- and experience accumulates, however, these cost differ- uted to respiratory diseases also peaked on the same day. ences are likely to diminish. The existing interindustry In all, a close correlation between concentration of sul- cost variation nevertheless implies potential savings fur dioxide, total deaths, and deaths resulting from re- from differentiation between producers on cost grounds. spiratory diseases (particularly for people over 65 or less than 1 year old) was observed statistically between July 25 and August 8. If the same study were carried out for The Benefits of Pollution Control: 1974 and 1975, when air pollution was worse, more Health Effects conclusive evidence on the relation of pollution to the deterioration of human health might have been found. The health effects of pollution are discussed in Lave and Seskin (1977) and in Smith (1976). Air Pollution and Mortality A regression technique, after Lave and Seskin (1977), Air Pollution and Human Health has been used to try to isolate a pollution-mortality Fernicola and Lima (1979) have evaluated the degree relation (see Thomas 1980 for details). It should be of exposure of the population of Sao Paulo to carbon emphasized at the outset that the purpose is to explore monoxide. To determine the carboxihemoglobin the existence of any significant association between the (COHb) content by the spectrophotometric method, 327 two phenomena in Sao Paulo rather than to derive the blood samples were collected from adults. Thirty sam exact effect of pollution on mortality. As will be clear, the ples for each control group were taken from residents of absence of an adequate data base limits the explanatory Embu-Gauu, an area considered to have a low carbon power of the equations. More generally, one should not monoxide concentration. The following significant re- expect to adequately capture relatively short-term pollu- suIts were obtained, tion-mortality relations, which must be based on long- term exposure to the effluents. Furthermore, these re- Concentration of COHb gressions measure the marginal effects of pollution on Group in blood (percent) mortality with other variables held constant, but in Traffic poicemen 63 2 reality a simultaneous explanation of the changes in Smokers 2.1 3 0.68 other variables as well as in pollution and mortality Bus drivers might be required. Smokers 4.6 _ 1.94 Annual cross-sectional data for 1977 for the thirty- Nonsmokers 1.6 _ 0.48 seven municipalities in the csp were analyzed with the Control use of linear multivariate regression analysis. The mor- Smokers 3.8 _ 1.74 tality rate (deaths per 10,000 population) is the depen- Nonsmokers 0.8 ± 0.21 dent variable. Death is recorded at the place of domicile 160 Vinod Thomas Table 13-2. Pollution Control Costs for Some High-Pollution Industries: Sdo Paulo and Rio de Janeiro, 1980 (1979 U.S. dollars) Metropolitan Sao Paulo Metropolitan Rio de Janeiro Particulate Particulate emissions .emissions Average Capital Capital Cost- collection Tons per cost per Tons per cost per benefit Industry efficiency year Percent ton year Percent ton ratio Nonmetallic Cement 98 38,917 28.39 67 25,500 20.69 69 1.0 Quarrying 95 31,467 22.95 222 64,300 52.17 124 13.6 Ceramic Clay 90 19,900 14.52 292 9,930 8.06 94 n.a. Gypsum lime 90 2,419 1.76 952 . . . ... ... n.a. Asphalt 98 10,745 7.84 358 21,588 17.52 169 17.9 Concrete 66 3,370 2.46 490 316 0.26 563 62.1 Glass 80* 785 0.57 8,577 ... ... n.a. n.a. Chemicals 80* 12,252 8.94 625 ... ... n.a. n.a. Metallurgic Steel 98 6,863 5.01 1,058 1,268 1.03 1,879 13.6 Iron foundries 98 10,364 7.56 1,860 343 0.28 6,467 45.6 Total or average 94 137,082 100.00 468 123,245 100.00 151 n.a. ... Zero or negligible. n.a. Not available. * Approximation. Note: Ratio of particulate emission, Sao Paulo/Rio, 1.11. Ratio of control cost, Sao Paulo/Rio, 3.10. Source: Based on Kowalczyk (1980). rather than of death, although it is not known how long Population density is an important determinant of the the person had lived there. Data on total mortality are incremental damages from a given amount of pollution. used first. Then, on the assumption that the pollution The age of the population clearly increases the mortality effect is longer term, children less than 1 year old and rate. Average personal income is expected to affect mor- less than 10 years old are excluded alternatively. tality inversely; estimated per capita value added by The independent variables that represent environ- municipalities, which is admittedly a crude proxy mental, physical, and socioeconomic characteristics are for the income of people living in those locations, had to pollution level, population density, average income be relied on. For the municipalities the availability of level, hospital beds per person, and the percentage of the medical service was approximated with data on hospital population 65 years and older.7 Since air quality data beds per person, as a socioeconomic variable. are not available at the municipal level, annual emission Equation 13-1 regresses the total mortality rate data were used instead. The emission data for particu- (deaths per 10,000 population) for the thirty-seven lates, sulfur dioxide, and carbon monoxide are available municipalities on total pollution (tons a square kilo- for the six municipalities which account for about 80 meter) and the socioeconomic variables for 1977. percent of industrial pollution. For the other municipal- ities levels of particulates and sulfur dioxide had to be (13-1) TM 76.214 + 0.120 PM - 0.068 S02 approximated with the use of estimated emission factors (1.51) P - 1.03) (emissions for a quantity of industrial output), and the level of carbon monoxide had to be estimated from an - 0.020 CO + 4.357P365 emission factor based on the number of vehicles. All the (- 0.67) (1.25) emission variables are expressed as a fraction of the area - 67.893 VA - 0.248 HB in square kilometers, to better reflect their concentra- 0.43) 1.46) tion. Since estimates of particulates are far more reliable than estimates of sulfur dioxide and carbon monoxide, - 0.002 Plkm2 + e the latter two pollutants are excluded in one set of (- 1.43) calculations. R2 =0.341 Evaluating Pollution Control: The Case of Sdo Paulo 161 where TM is the total mortality rate for the municipality, of the population over 65 is consistently and signifi- PM is mean particulate matter, SO2 iS mean sulfur diox- cantly related positively to the death rate. The variable ide, CO is mean carbon monoxide, P ¢65 is the percent- for per capita hospital beds also has a significant and age of the population aged 65 and older, VA is per capita negative effect on the mortality rate. Population density value added, HB is hospital beds per 1,000 people, and has a negative effect, which is contrary to expectation.9 Plkm2 is population density. The units of the variables That income is not a significant factor in any of the are given in table 13-3. The numbers in parentheses are equations is probably explained by the weakness of the t-statistics. data. Only 34 percent of the variation in the total mortality According to equation 13-1, an annual increase of 1 rate is explained by the seven independent variables ton of particulates per square kilometer in the GSP over (R2 = 0.341).' The exclusion of important socioeco- 1977 particulate levels is associated with an increase in nomic variables in the equation may partly account for the mortality rate of 12 deaths per million people. On the its low explanatory power. The noteworthy result is the basis of a mean particulate concentration of about 17 positive and significant coefficient for particulates. Be- tons per square kilometer" and a mean mortality rate of cause particulate data are the most reliable, this result is 8,830 per million, a 50 percent reduction in industrial encouraging. The other pollution variables, mean sulfur particulates alone" would be associated with a 1.2 per- dioxide and mean carbon dioxide, have no significant cent reduction in the mortality rate. Inclusion of better effect in the regressions, but the poor quality of the data estimates of mean sulfur dioxide and carbon monoxide may account for this result. As expected, the percentage levels may be expected to raise this effect significantly.'2 Table 13-3. Total Mortality per 10, 000 Population, Thirty-seven Municipalities, GSP, 1977 Particulates Particulates, SO2, CO Value TMT TM, TM9 TMT TM, TM9 R 2 0.175 0.343 0.366 0.341 0.451 0.470 Constant 69.079 22.290 25.618 76.214 27.913 30.759 Air pollution variables (tons a square kilometer a year) PM 0.006* 0.002 0.002 0.120* 0.075* 0.075* (1.423) (0.629) (0.584) (1.509) (1.630) (1.460) so2 0.068 -0.032 -0.030 (-1.031) (- 0.843) (-0.694) Co - - 0.020 -0.127 -0.017 (-0.667) (0.715) (-0.848) Socioeconomic variables P 65 (percent of total) 5.331* 7.449** 8.723** 4.357 6.633** 8.015** (1.606) (3.984) (4.140) (1.245) (3.290) (3.525) VA (cruzeiros per capita) - 189.581 14.330 52.335 -67.893 61.350 112.000 (-1.251) (0.168) (0.543) (-0.430) (0.674) (1.091) HB (per 1,000 population) -0.251* -0.166* -0.228* -0.248* -0.162* -0.225** (-1.393) (-1.640) (-1.996) (-1.458) (-1.547) (-2.037) P/km2 _ _ _ -0.002* -0.119* -0.001* (-1.431) (-1.599) (-1.432) - Not applicable. Notes: The numbers in parentheses are t-statistics. ' Significant at 10 percent level. *S Significant at 5 percent level. Variables: TMT = total mortality rate TM, = mortality rate excluding children 1 year old and younger TM9 = mortality rate excluding children 9 years old and younger PM = particulate matter So2 = sulfur dioxide CO = carbon monoxide P' 65 = population 65 years and older VA = value added HB = hospital beds P/kmn2 = population per square kilometer 162 Vinod Thomas With child mortality excluded, an increase of 1 ton of tion-an improvement by 1 microgram per cubic meter particulates per square kilometer may be associated with is associated with a decrease in mortality of 8 deaths per an increase of about 8 deaths per million. Given a mean 100,000-which may be explained by several factors. mortality rate of about 5,690 per million for the popula- The seven subdistricts face about the worst pollution tion over a year old, the implication is that a 50 percent problem, and marginal benefits from control at existing reduction in particulates alone is associated with an pollution levels are high. The pollution variable may also approximate 1.2 percent decrease in the mortality rate of be a proxy for excluded socioeconomic variables. The noninfants. relatively low level of public services in some of the more As a further check on the above results, data for seven polluted areas may contribute to higher mortality rates. highly polluted subdistricts within the Sao Paulo mu- Thus, the strong marginal mortality effect found should nicipality over the six-year period 1973-78 were not be fully attributed to pollution, but the consistently analyzed.'3 Data on total mortality for all age groups significant coefficient of particulate concentration may were available from Secretaria de Economia e Plane- be additional evidence of a significant incremental pollu- jamento (SEPLAN). Mean monthly pollution readings by tion-mortality effect in areas that are already heavily the CETESB, in micrograms per cubic meter for particu- polluted. lates, sulfur dioxide, and carbon monoxide, were aver- aged for 1973-78.' The weakest link in the data is income; per capita income was estimated rather roughly Policy Implications by EMPLASA for various zones which were aggregated to correspond to the subdistricts. It was indicated that if 25-50 percent of the GSP'S Equation 13-2 regresses the total mortality rate population is assumed to be directly affected by pollu- (deaths per 10,000 population) for the seven subdistricts tion, the per capita costs of a 94 percent abatement by during 1973-78 on mean and maximum particulate and the industries (or of a 55 percent reduction in total sulfur dioxide concentrations in micrograms per cubic particulate emissions in the area) would be $1.1-$2.2. A meter. It is assumed that family income grew by 3 per capita cost figure of about $4 is implied by Kowal- percent in real terms between 1973 and 1978. (Other czyk (1980) for the abatement of 71 percent of industrial results are in table 13-4.) particulates in the heavily industrialized Medio Paraiba area, on the assumption that only 75 percent of the (13-2) TM' = - 134.99 + 0.888 PM mean people are directly hurt. It was also suggested (with (5.72) caveats) that air pollution is statistically significant in + 0.153 SO2 + 10.655 Pr 60 explaining variations in the total death rate in the GSP. (1.27) (4.28) Although no dollar value was placed on the health benefits associated with pollution control, the orders of + 0.009I magnitude of costs of control suggest that net benefits (2.83)5 could possibly be obtained from a substantial reduc- - 0.003 Plkm2 + e tion-say 85-90 percent-in the stock of pollution. In (4.76) R2 = 0.579 the GSP smoke abatement can be efficiently achieved by focusing on a relatively few large polluters, and this is where TM' is the subdistrict mortality rate, PM mean the current policy. Concentration on the large nonme- and S02 mean are the average particulate and sulfur tallic industrial polluters may also make sense on eco- dioxide concentrations, respectively, P ¢ 60 is the per- nomic grounds, but such a discriminatory approach may centage of the population aged 60 or older, I is per capita not be practical. income, and P/km2 is population density. In table 13-4 Of the methods of reducing pollution, the strategy of PM max and S02 max are annual averages of monthly affecting industrial output is relatively inefficient be- maximum values of particulate and sulfur dioxide con- cause it does not lead to substitutions in the production centrations. processes which could lower the cost of controls. The About 60 percent of the variation in mortality among preferred approach would be to directly control pollu- the subdistricts is explained in equation 13-2. Inclusion tion either by an emission tax or by pollution abatement of PM max and SO2 max raises the explanatory power to standards, which induce producers to cut back their use 66 percent." A change in the assumption of constant of pollutant fuels or find substitutes for them and to income to a 3 percent growth in income has little effect. adopt control equipment. Direct fuel restrictions pro- Similarly, the deletion of the variables PM max, SO2 vide no incentives for the use of control equipment. max, and Plkm2 does not affect principal results Similarly, requirements to use abatement equipment appreciably. give no inducements to restrict or modify fuel use. Equation 13-2 implies a high and significant associa- The higher intensity of industrial activity and the Evaluating Pollution Control: The Case of Sdo Paulo 163 Table 134. Total Mortality per 10, 000 Population for Seven Subdistricts, Sdo Paulo Municipalities, 1973-78 Constant 3 percent annual growth real income in real income Value A B C A B C R2 0.317 0.587 0.667 0.314 0.579 0.663 Constant -38.330 - 136.380 - 124.025 -37.548 - 134.990 - 122.236 Air pollution variables (micrograms per cubic meter) PM mean 0.472** 0.894** 1.097 0.469** 0.888** 1.119** (2.94) (5.81) (5.44) (2.91) (5.72) (5.48) SO2 mean 0.040 0.160* 0.062 0.033 0.153 0.065 (0.27) (1.34) (0.493) (0.225) (1.27) (0.51) PM max - - -0.220** - - -0.235 (-2.25) (-2.38) S02 max - - 0.147* - - 0.134 (1.33) (1.21) Socioeconomic variables P 60 (percent of total) 3.312* 10.55* 9.112** 3.499* 10.655 9.209* (1.33) (4.28) (3.87) (1.40) (4.28) (3.89) Per capita income (cruzeiros) 0.003 0.009 0.010** 0.002 0.009** 0.010** (0.78) (2.97) (3.53) (0.67) (2.83) (3.44) P/km2 - 0.003 -0.003** - - 0.003** -0.003** (-4.85) (-5.42) (-4.76) (-5.37) - Not applicable. Notes: Numbers in parentheses are t-statistics. Variables are defined in table 13-3. Columns A: variables PMmax, SO2 max, andP/km2 excluded; columns B: variables PM max and SO2 max excluded; columns C: all variables included. * Significant at 10 percent level. ** Significant at 5 percent level. greater population density in Sao Paulo, combined with asphalt plants is primarily attributable to the inclusion the city's unfavorable geographic and climatic charac- of fugitive dust emissions in the Rio inventory. Ceramic teristics, imply higher incremental damages there than plants in Rio, which are small in size, can use water in Rio. According to the 1970 Industrial Census, the spray systems instead of more expensive bag filters. Rio's value of production in Sao Paulo was 8.6 times that in higher cost per ton for foundries and steel mills appears Rio. The difference in particulate effluents indicated in to be a result of smaller-size facilities, for which the cost Kowalczyk (1980) for 1979 is only 11 percent, which per ton of initial abatement hardware is higher. appears to be the result of a serious underestimation of Thus, while higher apparent pollution damages in Sao fugitive dust in Sao Paulo. When fugitive dust sources Paulo call for more stringent abatement than in Rio, the from quarries and asphalt plants are excluded, Sao generally higher range of control costs in Sao Paulo Paulo's effluents exceed Rio's by 145 percent. Given Sao limits the desirability of controls. In the absence of Paulo's larger area, however, pollution concentration in better estimates of benefits and costs, one cannot con- the two places would not be of the same order of magni- clude whether it would be socially beneficial to treat Sao tude. Population density in Sao Paulo exceeds that in Rio Paulo and Rio differently under an antipollution (by about 17 percent in 1970 and more at present), strategy. More information would make possible a cost- which would lead to a significantly higher marginal benefit ranking of the various control options for each benefit of pollution control in Sao Paulo. area, with implications for differential treatment, to aid Rio's industries seem to face lower control costs than in identifying the proper priorities. those in Sao Paulo (table 13-2). With quarries and Much greater differences in damages from pollution asphalt plants included, the difference in cost per ton of in the Sao Paulo-Rio comparison may be expected if pollution control between the two places is 210 percent. smaller towns are also considered. The desirable ap- Because of the inaccuracy of data on fugitive dust, proach would be to set widely different abatement goals however, a better comparison is provided by excluding for different locations. Two other strategies may be con- quarries and asphalt plants; then the cost difference is 60 sidered: one that requires uniform percentage abate- percent. The large difference in the control cost for ment in all areas and another that sets equal maximum 164 Vinod Thomas allowable emission standards for all places. In principle 7. Data sources are Secretaria de Economia e Planejamento Brazil has uniform emission standards; in practice they (SEPLAN); Funda,co Sistema Estadual de Analises de Dados are enforced only selectively and through more or less Estatisticos (SEADE); Secretaria de Estado de Saude, Centro de uniform percentage abatement requirements. The re- Informacoes de Saude (cis); and Secretaria de Fazenda. sulting spatial differentiation, although not necessarily the explanatory power to nearltality and child mortality raises ideal, would be better than that implied by the imple- particulates falls, however, presumably because of the vulner- mentation of identical emission standards. ability of babies to pollution. The least amount of spatial shifts in production would 9. This may suggest that density is endogenous within the be induced by the uniform abatement (percentage col- model in adjusting for pollution levels (that is, the impact of lection) policy. If, in addition, new sources are singled particulate concentration on density is negative), and thus out for regulation, or if they face significantly stricter density has a negative relation to mortality. restrictions, the existing locational patterns would tend 10. About 137,000 tons of annual particulate emissions by to be frozen. In contrast, the desirable policy would industries is divided by the GSP'S area of 7,951 square kilo- normally require more abatement in the more damaged meters. places and thereby encourage some displacement of 11. This means about a 33 percent reduction in total par- existing output and induce future entrants away. An ticulate emissions, since industries contribute 65 percent. existing output unrean since fussibitu s etrants awa y. , 12. Lave and Seskin (1977) found that for 117 Standard extreme but not unrealistlc posslyity iS that an opbtmal Metropolitan Statistical Areas (sMsAs) in the United States for policy would concern itself only with highly built-up 1960 a 50 percent reduction in particulates and sulfates was urban centers and would set no controls on highly rural associated with a 4.7 percent decrease in the mortality rate. situations. As a result industrial plants would be induced 13. The subdistricts are Aclimacao, Cerqueira Cesar, Con- to move away from population centers, which would solac&o, Indianapolis, Lapa, Santa Cecilia, and Tatuape. help to cut down on the total cost of antipollution 14. Other data sources were SEPLAN, SEADE, cis, and policies. EMPLASA. 15. The negative coefficient of PM max is unexpected and is contrary to findings reported earlier by Mendes (1976). Notes 1. On the basis of projections in Hauser and Gardner (1982) and United Nations (1980). Problems arising from conflicting Bibliography definitions of boundaries make comparisons difficult. Today Tokyo and New York would also be larger than Sao Paulo if Companhia Estadual de Technologia de Saneamento Ambien- Yokohama and northeastern New Jersey were included in the tal (CETESB). 1979a. Legisla,do basica polui,do ambiental respective metropolitan areas. estadual e federal. Sao Paulo. 2. In New York City and Chicago, for instance, sulfur diox- - . 1979b. Legislacao basica regioes metropolitanas pro- ide concentration averaged 150-160 micrograms per cubic te,do dos mananciais e zoneamento industrial. Sao Paulo. meter and suspended particulates 180-190 micrograms per Empresa Metropolitana de Planejamento de sao Paulo cubic meter in 1972. The average carbon monoxide concentra- (EMPiAsA). 1979. Zoneamento industrial metropolitano, a tion recorded in a representative area in New York City in 1975 experiencia da Grande Sao Paulo. Sao Paulo. was 4.0 parts per million. Esteves, Silvio S., and Adenir Gianneschi. 1980. "O sistema do 3. The welfare cost of pollution control, as in Thomas (1980, inventario de fontes de poluico ambiental como suporte no 1982), would be somewhat lower than these measures. controle de poluisdo do ar na regiao metropolitana de Sao 4. The Sao Paulo municipality, parts of which are heavily Paulo." International Clean Air Congress, Buenos Aires. polluted, contains over 65 percent of the GSP'S population. The Processed heavily industrialized areas of Santo Andre, Sao Caetano, Sao Pe d. Bemardo, Osasco, Mogi das Cruzes, Guarulhos, and Diadema Fericola, Nilda G. G., and Fausto Azevedo. 1979. "Niveis de account for another 20 percent. chumbo e atividades do acido daminlevulene-codesedratasa 5. A 1980 population estimate for the industrialized muni- no sangue da populaqdo de Sao Paulo (Brasil)." Presented at cipalities that constitute this area is 1.16 million, or a popula- the 10th Congresso Brasileiro de Engenharia Sanitaria e tion density of 224 persons per square kilometer. Ambiental, Manaus, January 21-26. Processed. 6. It does not follow, however, that to meet overall abate- Femicola, Nilda G. G., and Ernesto R. Lima. 1979. "Avaliacao ment standards which are significantly higher than 85 percent do grau de exposicao de amostras populacionais de Sao it would be desirable to raise the standards for the larger Paulo ao monoxido de carbono." Rev. SaudePublica, no. 13. producers above 94 percent. Instead, as the overall standards Sao Paulo. are raised, it would most likely be desirable to include the Friedlaender, Ann F., ed. 1979. Approaches to Controlling Air smaller polluters in the operation. Pollution. Cambridge, Mass.: MIT Press. Evaluating Pollution Control: The Case of Sao Paulo 165 Gianneschi, A., A. P. Junior, and N. N. B. Salvador. 1979. Pazzaglini, Filho M., and M. A. Greco. 1981. Zoneamento "Inventario das fontes de poluic&o no estado de SAo Paulo." industrial em areas criticas de poluiqio. SaO Paulo, CNDU, Trabalhos da CETESB, vol. 1. S&o Paulo. Ministry of the Interior, State of Sao Paulo. Hauser, Philip M., and Robert E. Gardner. 1982. "Urban Fu- Smith, V. K. 1976. The Consequences of Air Pollution. Cam- ture: Trends and Prospects." In Philip M. Hauser and others, bridge, Mass.: Ballinger. eds., Population and the Urban Future. Albany, N.Y.: State Thomas, Vinod. 1980. "Welfare Cost of Pollution Control." University of New York. Joumal of Environmental Economics and Management, Kowalczyk, J. F. 1980. "Rio de Janeiro Industrial Pollution vol. 7, pp. 90-102. Control Project." LatinAmericaand the Caribbean, Country . 1981. Pollution Control in Sao Paulo, Brazil: Costs, Programs Department, World Bank, Washington, D.C. Pro- Benefits, and Effects on Industrial Location. World Bank cessed. Staff Working Paper 501. Washington, D.C. Lave, Lester B., and Eugene P. Seskin. 1977. Air Pollution and _ . 1982. 'Welfare Analysis of Pollution Control with Human Health. Resources for the Future Series. Baltimore, Spatial Alternatives." In George S. Tolley and Philip Graves, Md.: Johns Hopkins University Press. eds., Environmental Policy. vol. 2:Air Quality. Cambridge, Licco, E. A., H. I. Oda, and J. B. GalvAo Filho. 1979. "Padroes de Mass.: Ballinger. emissao para controle de material particulado na regiao da _ . 1985. "Evaluating Pollution Control: The Case of Sao Grande Sao Paulo." Trabalhos da CETESB. Sao Paulo. Paulo, Brazil." Joumal ofDevelopment Economics, vol. 19, Mendes, Rene. 1976. "Avaliacao de efeitos da poluicao do ar pp. 133-46. sobre a saude, atraves do estudo da mortalidade diaria na Tolley, George S., Philip E. Graves, and John L. Gardner, eds. GrandeSdoPaulo-1973."cETESB Project36176,part 1.Sao 1979. Urban Growth Policy in a Market Economy. New Paulo. York: Academic Press. Mills, Edwin S. 1978. The Economics of Environmental Qual- United Nations. 1980. Pattems of Urban and Rural Population ity. New York: W. W. Norton. Growth. Population Studies 68, table 4-7. New York. 14 Improving the Effectiveness of Urban Projects Douglas H. Keare E arlier chapters have reviewed the urban policy op- pilot program to rigorously evaluate selected early proj- Etions that are available to developing countries. A ects. Preparation for the evaluation began with approval variety of policies intended to alleviate urban problems of the first project in 1972, but it was mid-1975 before and improve the functioning of cities have been used or enough projects were ready for implementation. The proposed for use by developing countries. The results to housing program was launched with the assistance of date have been mixed and controversial, and there is an the International Development Research Centre (IDRC) emerging consensus that better evaluation of the actual of Canada. Programs in Senegal (the first project), El and possible outcomes of urban programs would be Salvador, the Philippines, and Zambia were selected for helpful to policymakers who must choose between poli- study, and annual conferences were held to discuss the cies. This chapter examines whether rigorous evaluation findings. The concluding conference for this phase of of projects can assist in improving the efficiency and the work, held in November 1980 in Washington, D.C., effectiveness of future urban policymaking and in for- was attended by project managers and researchers from mulating and implementing projects. Although much of the countries involved and from other interested coun- the chapter deals with specific shelter projects financed tries. Subsequently the evaluators' attention has turned by the World Bank, the lessons drawn from such cases to the publication and dissemination of the programs' are likely to apply to policymaking as a whole. more important results and to evaluations of projects in In 1969 the World Bank began a lending program that other countries. was explicitly designed to respond to the compounding This chapter argues that evaluation programs can do problems of cities in developing countries. The emphasis much to guide public sector spending (or lending) pro- was on ameliorating the housing problems that faced grams. This was not the primary aim of this evaluation the burgeoning low-income populations in the cities by program at its outset; then the emphasis was on measur- accelerating the rate of increase in the supply of basic ing a broad range of expected impacts on the partici- shelter and by facilitating access to the expanded supply pant populations.2 As the program progressed, however, by the lowest income groups. The strategy developed and initial experiences and evaluation findings were was to provide secure tenure and basic services in new digested, the focus narrowed to issues that were seen to sites and services areas and in slum and squatter areas be both significant and amenable to rigorous evaluation designated for upgrading and to rely on the participants or research. As a consequence, although the study of to complete or improve their houses through progres- impacts has remained a central concem, it has been sive development, with a modicum of additional joined by the assessment of the effectiveness of pro- assistance.' The first such project was approved in June grams and projects and of certain of their components. 1972. Since then a sizable lending program has de- Ex post assessment of experience makes clear that veloped and has accounted for over $2 billion in loans for there can be a synergism of activities in a jointly pro- sixty-two projects through 1981. grammed sequence of projects. These activities are an An early feature of this unprecedented endeavor was a evaluation component, policy reviews at various critical 166 Improvizg the Effectiveness of Urban Projects 167 junctures, and associated research. As has happened in research than could have occurred through direct this program, project experience, evaluation findings, assessment of project experience alone. It concludes and research results all inform each stage of project with some observations on how the efficiency of interac- design and each policy review. At the same time, project tion between evaluation and research on the one hand and policy requirements, coupled with evaluation re- and project design and policy formulation on the other search experience, can help to adjust the designs of can be improved. evaluation and research programs to maintain respon- siveness to these operational needs. The following ma- trix provides a helpful way of viewing this interaction. Project Efficiency and Effectiveness Activity Despite differences in the goals and operational proce- Continuous Periodic dures of the four shelter programs studied, the pro- Action Project Policy grams are comparable enough to warrant preliminary Study Evaluation Research gasaecmaal nuht arn rlmnr investigation of their efficiency and effectiveness accord- The general background against which effectiveness is ing to a single set of criteria. Eight project aspects reviewed in this chapter is one of success. The progres- appear to be of particular financial, economic, and social sive development model, exemplified by the sites and importance in all four cases: project planning and de- services and area upgrading projects, has demonstrated sign, selection of project beneficiaries, construction its validity in a broad range of circumstances, and the methods, materials loan programs, housing completion recorded impacts on the participating populations and and occupancy, maintenance of housing and infrastruc- their housing conditions have been in the expected ture, cost recovery, and community participation. directions and of significant magnitude. Self-help con- Criteria that appear to be useful in gauging the efficiency struction methods have proved relatively efficient, and of these components include speed of implementation, the impacts of projects on the housing stock have been cost, quality of housing or services, accessibility by generally greater than anticipated. The projects have target populations, replicability, and flexibility of imple- been affordable and generally accessible for the target mentation. populations. Although it is still too early for some of the There are tradeoffs among these evaluative criteria. developmental results to have been recorded, those Higher-quality housing may be more costly to construct measurements which have been completed indicate that and less affordable for low-income target groups. the projects' impacts on the socioeconomic conditions Greater attention to selection procedures may increase of participants have been in the directions expected. target group accessibility but raise the costs of identify- And, notably, the projects have not had negative impacts ing appropriate families. The sometimes conflicting on expenditures for food and other basic necessities objectives have to be reconciled according to a crude (see Bamberger, Gonzalez-Polio, and Sae-Hau 1982; relative weighting of each, as determined by project Bamberger, Sanyal, and Valverde 1982; Keare and Ji- management and other policymakers, and no precise menez 1983; Keare and Parris 1982). model for the assessment is yet available. The discussion Notwithstanding this general record of success, the below briefly summarizes the constraints and options projects have encountered some problems and produced involved in applying the criteria to each project aspect some unexpected results, some of which are detailed in and reports on selected initial results from the four the following section. An analysis of the projects' suc- projects. cesses and shortcomings supports recommendations, advanced below, that future projects endeavor to push Project Planning and Design standards and costs still lower, include explicit provi- sions and opportunities for rental arrangements, and Speed of implementation can affect the cost of both incorporate credit provisions that are more nearly tai- house construction and provision of services. If progres- lored to the needs of the targeted families. Furthermore, sive development is too slow administrative costs per it is observed that these objectives will be more easily plot will rise and intended benefits will be delayed and pursued and also more effective if a rigorous and relent- possibly reduced, but pressure to construct houses too less approach is taken to improving cost recovery per- rapidly, without appropriate credit, can result in afforda- formance in the programs. The final section, "The Con- bility problems for families. High, rigidly enforced de- tributions of Evaluation," provides some examples of sign standards for quality housing are inexorably linked how operationally useful knowledge and understanding to higher costs, may contribute to delays, and may limit have been developed more rapidly, completely, and con- participation by low-income families and restrict the vincingly through the use of evaluation and associated flexibility of households' responses. Reductions in capi- 168 Douglas H. Keare tal costs, however, may produce only illusory gains be- (for example, by producing their own soil-cement cause they are offset (or more than offset) by increases in blocks), thereby reducing costs and potentially increas- operating and maintenance costs. Although costs can ing long-term efficiency, have so far lacked appeal. In sometimes be reduced by the use of domestic rather Senegal a similar desire for additional space rather than than imported materials, families may prefer to build highly serviced facilities has been demonstrated by the with more expensive materials and recoup the additional choices of the low-income control groups in Guediawaye outlays through higher returns from renting part of the and Grand Yoff. The relatively expensive standard de- space and from eventual resale. signs were presented to the project families in such a There is no evidence that design standards in the way that the costs, in time and money, of obtaining projects have been intolerably high. Studies of turnover approval for simpler designs were perceived to be sub- do not, for example, reveal a higher than average inci- stantial. This belief may well have led to families' accept- dence of departure from project sites among the poorer ing housing standards that were higher than desired. participants.3 Two facts, however, seem clear: entry These findings point to the advisability of providing a costs can be reduced so as to facilitate accessibility for wider range of building options to participant families. poorer households without adverse consequences for Households appear to demand substantially different physical results, and, even if costs remain the same, combinations of plot size, contractor construction, and changes can be made in the mix of components and service facilities; they also wish to have the option of services offered that would make them more desirable. adding to or further upgrading their homes and of rent- In particular, the evaluation program and other studies ing out space. Despite the counterarguments of some have shown that the projects under review have placed national housing planners, it appears that scaled-down somewhat too high a premium on service levels and may housing standards are a precondition for affordable (but have overlooked opportunities to facilitate increases in still desirable) housing for low-income populations. housing space-an attribute that has turned out to be Rental arrangements, too, appear to be both desirable more highly valued than was anticipated in the initial and feasible for target populations, although strict en- project designs. Although families in El Salvador have forcement of cost recovery is mandatory to avoid the on the whole been satisfied with the design features of creation of undue subsidy of rentier groups within the project housing, demand studies have revealed that they projects. These issues will be treated below. also view favorably the larger lot sizes (and lower service levels) of colonias ilegales (unauthorized settlements) if this option is available in a particular city.4 Similarly, in the Philippines, where project lot sizes in the Tondo area The selection procedure also affects the efficiency of of Manila have been constricted by the existing high sites and services programs.6 Project managers are typi- densities, families have responded, following the re- cally concerned about both the upper and the lower blocking process, by building additional stories to in- income bounds for selections from the applicant pool. crease living space.5 Improved services, especially toilet Care must be taken to avoid including in the project too facilities, have been less highly valued by upgrading many high-income families, some of whom may attempt families than had been anticipated by project designers. to falsify income statistics to participate, but it is also In Zambia the plot sizes offered have been large necessary to ensure that low-income participants will be enough to satisfy families, but apparently even larger able to afford project costs. The speed of the selection lots would have been preferred. A partial explanation is process itself involves tradeoffs. If the screening pro- that families have discovered that they can rent out gram takes too long, households may become dissatis- space despite project prohibitions. The outcome, how- fied and withdraw from consideration; if it is carried out ever, is biased by the fact that Zambian participants are too rapidly, errors may occur. not charged for land-which has potentially serious Restricted definitions of income may also lead to consequences both for project costs and replicability and biased selection standards. If selection is based only on for the allocation of a scarce factor of production, land. earned income rather than on total income from all The primitive status of the building materials industry sources, projects may exclude large numbers of house- and the excess demand for good housing in Zambia help holds which, by virtue of the combined resources within explain Zambian families' preferences for constructing their extended families, can afford the programs. Fur- houses of imported rather than domestic materials: they thermore, the criterion will be biased against female- incur added costs for higher quality and recoup those headed households, which rely more on transfer pay- costs by renting out space. In these circumstances ments from kin than do male-headed households. It efforts to encourage participants to use local materials must be noted, however, that verification of the true Improving the Effectiveness of Urban Projects 169 incomes of all applicants is a costly procedure. Statisti- and supervise individual workers; and self-help in which cal sampling procedures will probably have to be used families build their homes themselves. Each method and the inherent errors accepted. involves different construction rates, costs, and implica- The evaluation studies have not discovered seriously tions for housing quality, replicability, and accessibility, anomalous income distributions in any of the four proj- and the tradeoffs involved are considerable. The mutual ects. An acceptable range of income groups appears to help method may be slower (although not necessarily have been incorporated into the programs.' Some prob- less expensive) than certain self-help methods, which lems with underreporting of incomes arose in the Dakar may in turn be slower (but cheaper) than contractor project, but a more general finding has been that some construction. The scheduled timing for mutual help families, particularly female-headed households and methods may either attract or discourage families with those employed in the informal sector, initially suffered different preferences and varying formal and informal from a measure of unintended discrimination in the work schedules; scheduling is a serious issue. Use of selection stage, owing to problems in verifying income skilled builders can raise project costs in the short run, from nonformal sources.8 Several of the El Salvador but their participation in project design and execution subprojects have gone so far as to include higher propor- may also contribute to replicable plans which scale down tions of female-headed families than exist in the popula- costs, including maintenance costs, over time. tion at large. All four country projects have included Family-based construction methods using progres- families with a greater range of incomes (and more sive development have proved viable in sites and services diverse sources of those incomes) than had been fore- and upgrading projects. It has been estimated that cer- seen. Consequently, future programming efforts may tain families in El Salvador have saved up to 30 percent well have to find a way to assess both household and of costs by building their own houses rather than hiring extrahousehold incomes and characteristics to deter- contractors, and these houses have been judged to be of mine an appropriate range of income bounds for project a quality comparable to those constructed by skilled participation, rather than necessarily modifying the builders. Combinations of self-help and mutual help can selection processes themselves. also produce housing of acceptable quality at costs simi- The evaluations have demonstrated that transfers lar to or lower than those charged by contractors. Yet from the extended family and other kin and nonkin such methods can absorb great amounts of supervision networks account for an important and stable share of and elapsed time. In El Salvador the mutual help phase participant families' incomes. In the sample of low- of the first projects, although it produced good results, income households in Santa Ana, El Salvador, transfers lasted more than forty weeks, since work was carried out constituted 66 percent of the incomes of 58 percent of only on weekends. The original design also underesti- the families in the lowest income decile; for the next mated the opportunity costs of labor in general, and it lowest decile the transfers were 25 percent of the income was found that skilled workers and other small business- of 48 percent of the families. In the Tondo sample trans- men, in particular, could have been much more profit- fers accounted for 54 percent of the incomes of 43 ably employed elsewhere during these work periods.9 percent of the families in the lowest third of the income Furthermore, households headed by single adults and distribution. Furthermore, econometric analyses indi- especially by females typically found participation both cate that transfer incomes have a contractual character undesirable and difficult, owing to several constraints and are devoted more than proportionately to expendi- on the flexibility of their time. Sweat equity does not tures which can be described as meeting basic needs. appear to have been available for own construction in Participation in projects (which confers investment the relatively abundant quantities assumed, and as many opportunities) induces additional transfers (Kaufmann as 51 percent of the households in the Santa Ana and 1982). Sonsonate projects used only hired labor to bring their core housing up to habitable levels. (See table 14-1 for a Construction Methods comparison of housing construction methods in El Sal- vador.) The projects have placed substantial premiums on These findings imply that market solutions for hous- identifying construction methods that reduce costs ing construction, such as seeking income transfers, hir- and contribute to efficient implementation. Among the ing labor, and balancing accounts through the house- cost-efficient methods are project-provided, contractor- hold's own greater participation in the labor force and built housing; construction through mutual help; self- rental of part of the structure, ought to be considered in help in which families hire a contractor; self-help in the estimation of housing costs and other aspects of which families themselves do the contracting and hire project design in future shelter programs. Pure self- 170 Douglas H. Keare Table 14-1. Housing Construction Methods in El Salvador: Santa Ana and Sonsonate Unpaid Hired and Hired labor family unpaid family All Item only labor only' labor options Percentage of participants 51 27 21 100 Average number of person- days of hired labor 46.0 - 33.2 42.4 Average weekly wage income (colones) 124.0 120.0 155.0 130.0 Average number of household members with construction experience 0.03 0.41 0.12 0.16 Average number of person- days of unpaid labor - 44.3 45.0 44.6 Average weekly nonwage income (colones) 14.0 9.0 6.0 10.8 -Not applicable. Note: Figures may not add to totals because of rounding. a. Regression analysis indicates that wage rate and proxy variables (such as technical experience, number of adults ages 17 to 60, and the male-female mix within a household) best measure a household's ability to build its own dwelling. There is also some evidence thathouseholds that use self-help have lower elasticities of housing demand than others. Source: Data from Fundaci6n Salvadorefia de Desarrollo y Vivienda Minima. help, it seems, should be envisioned as only one of a set the materials are too costly or too difficult to acquire and of diverse methods that can be employed for efficient distribute. housing construction and consolidation. As the El Salva- In Zambia problems in stocking on-site stores led to dor data demonstrate, the amounts of self-construction considerable delays in distributing materials, and fami- undertaken will depend, for each household, on the lies queued for as many as twenty hours a week to receive productivity and the opportunity costs of the family items. Thefts of materials also occurred. members who engage in the associated activities and Restriction of purchases of materials to project stores hence, on the level and sources of household income and may have compelled families to buy higher-quality sup- the family members' construction skills, among other plies than necessary or desired. It appears that in the El factors. Pepito and San Jose del Pino projects in El Salvador materials could have been purchased directly at Materials Loan Progranzs wholesale prices, thus lessening or obliterating the ex- pected cost-reducing advantages of the stores' bulk In the projects under consideration, loans for housing purchases. construction were limited to credits for materials. In the Philippines the project stores did not stock the Efficiency evaluations must thus gauge whether the cement, hollow blocks, or steel bars most desired by materials purchased have been appropriate and afford- households. Tondo dwellers, on the whole, appear to able for housing consolidation and have been distributed have found cheaper (but acceptable) materials else- speedily enough to participants, and whether other where, with the Housing Materials Loans Program forms of housing loans might have been advisable. The (HMLP) providing only 25 percent of beneficiaries with credits have frequently been tied to purchase of mate- construction materials. rials from project stores. Although such approaches are The credit programs themselves have been deficient designed to compensate for supply failures and to cap- in certain respects. In the Philippines, Senegal, and ture for participants the savings inherent in bulk pur- Zambia, for example, the maximum sizes of loans avail- chases, newly created project stores may have high able were insufficient to cover the requirements of many administrative costs, encourage overly high building households that sought credit to finance construction. standards, and fail to recommend local materials which, Total funds have sometimes been exhausted before all though acceptable for use, do not fall under the stores' requests have been met.'" In the face of grossly inade- control. Although sheer mechanical replicability may quate information on the demand for credit, it is under- appear to be facilitated through this relatively simple standable that project designers were loath to encourage tied credit option, the technique can be compromised if families' overextending themselves financially and Improving the Effectiveness of Urban Projects 171 sought to avoid this outcome by restricting both the size to obtain land entitlement rights. These delays, and of individual loans and the total funds available. It others outside project control, have exacerbated the appears, however, that these restrictions had undesir- already difficult problems of coordination with other able effects on some participants."1 agencies responsible for water, lighting, and other basic Notwithstanding these difficulties, studies to date services. The National Housing Authority in the Philip- have concluded that credit programs have been useful pines is obliged to negotiate with no fewer than eleven and that families should simply be permitted greater other agencies to deliver inputs in a mutually com- flexibility in seeking housing materials in the market plementary fashion. (except, perhaps, in countries such as Zambia where Delays in the installation and effective functioning of materials are in acutely short supply). The empirical fact services may also delay house consolidation and plot that project families hire labor for a substantial propor- occupancy.13 Families in turn are confronted with decid- tion of housing construction argues further for consid- ing when, how, and in what sequence to construct the eration of more general lending, where appropriate, that house and move to the plot. If they can move to the plot would cover the costs of labor, particularly skilled build- immediately and commence construction, there will be ing help, as well as that of materials. Some scholars no problem for the individual family."4 Several factors contend, moreover, that the substitutability of labor for may, however, delay the move and force families into a materials in housing construction may be limited, so situation in which they are simultaneously paying rent that tying credit to materials generally may pose con- on their current residence and charges on their new straints which are only slightly less severe than tying to plot. The two principal such factors are lack of services materials from specific stores.'2 and the absence of a habitable structure on the site. Little evidence exists as yet in the economic, financial, Others are distance of the new site from the old (and or even sociological literature which would offer guid- from the workplace) and inadequate credit." World Bank ance on the optimum forms and amounts of credit. It is project designers responded quickly to this situation by evident that research into both demand for credit and adding a measure of core construction to most projects. the responses of credit markets is a prime necessity. Doubtless this method is one way of effecting earlier Studies now under way in the World Bank stress inves- occupancy and avoiding certain costs, but it also adds, tigation of informal credit markets, particularly their sometimes substantially, to direct costs and is generally linkages with household expenditures for basic needs, a less flexible option. An urgent need of the lending and this emphasis should enable the research to contrib- program is to evaluate the tradeoffs between this and ute to improvements in the design of urban shelter more flexible approaches which may be sustained by projects. more innovative credit mechanisms. Housing Completion and Occupancy Mamtenance of Housmg and Infrastructure Occupancy of plots in sites and services projects is Maintenance of infrastructure facilities and housing another factor to be assessed in determining the effi- must also be designed and managed efficiently. Because ciency of project implementation. Unless sites are occu- these project features are often the responsibility of pied relatively rapidly, costs can escalate through added overburdened and underfunded local government agen- interest incurred during construction, through the real cies, there is a potential problem of inadequate mainte- effects of inflation, and through the expense for families nance. A related issue is the nature of maintenance of maintaining one residence while another is being costs: if these costs are too high, whether because of low built. Arguments for replicability are thus likely to be capital investment or poor design, the benefits of low- vitiated by slow inhabitation of project areas, since con- cost housing can be rather quickly dissipated; if design struction delays may cause affordability problems for standards are kept unreasonably low and do not facilitate poorer families and induce project managers to include maintenance, project facilities will decline in quality. more high-income households than originally desired to Though it is not entirely clear how the assignment of achieve full occupancy within a reasonable period. responsibility for maintenance might affect access to To date, land acquisition and the installation of basic projects for certain groups (such as higher-risk house- services have proved to be the most serious hindrances holds with less steady incomes), it is apparent that in- to speedy project implementation in the early (infra- adequate planning for maintenance can negatively affect structure) stages. Whether because of difficulties in government decisions regarding replicability. alienating public land, as in Zambia, or in finding afford- Because the evaluated projects have been installed for able and accessible terrain for lots, as in El Salvador, only two years or less, definitive statements on project project managers have had to struggle literally for years maintenance cannot be offered. Yet problems with gar- 172 Douglas H. Keare bage collection have already arisen in the Lusaka proj- assumption was that affordability problems must have ect. At first view the cause appeared to be difficulties in contributed to the result, but evaluation studies have servicing the garbage trucks and the resulting deteriora- established that repayment performance has not varied tion of vehicles, which led to dissatisfaction with the with income. The principal factor in the slow rate of trash collection services and negatively affected cost collections has been the absence of political will on the repayments by families. Further investigation showed part of the responsible political party to seek repay- that families were not totally satisfied with the services ments. The party did little to prompt communities to even when they were working-they wanted the services repay development costs to the Lusaka City Council brought closer to their houses. These doubts led to other (LCC), the project's administrative agency. The LCc has questions about the capacities of internal roads, and now taken steps to recover costs more systematically, hence about overall costs and charges. Project experi- but efforts are hindered by the absence of an efficient ence and evaluation have verified that these various recordkeeping system to maintain up-to-date informa- features-design, maintenance, cost recovery, and com- tion on defaulters and amounts owed and by the lack of a munity participation-are inextricably related and, flexible mechanism for collecting payments. The under- although it is difficult, better means of dealing with lying problem is the long-standing tradition of subsi- them simultaneously must be sought."6 Because nearly dized housing in Zambia, the potentially harmful effect all project services will require maintenance, planners of which was inadequately evaluated during project de- will have to take additional steps in the future to ensure sign and appraisal. Families, moreover, have not been that both cities and their inhabitants are financially and well informed about repayment schedules or about the institutionally prepared to continue maintaining facili- purposes for which the money collected was to be used."7 ties over time, especially in the light of the cost recovery Collections were also slow to develop in the early problems already encountered. projects in the Philippines. The initial investigation found that the poor early performance did not stem from Cost Recovery inability or unwillingness to pay or from poor collection procedures. Rather, the cost recovery objective was in- Given the desirable scope and scale of urban shelter sufficiently integrated into project planning, imple- programs, cost recovery has since the inception of these mentation, and monitoring procedures. Thus, project programs been deemed a crucial feature of project de- activities have been articulated not in a critical path sign because inability to collect payments would doubt- framework but rather in a serial framework. Typically, less prejudice replicability. Delays or dissatisfaction with when physical implementation has been completed and project services can make cost recovery more difficult, collections could ostensibly begin, some neglected and failure to recoup payments may in turn raise the feature has delayed collection: titling has not been com- costs of the project for executing agencies and for those pleted (or even started), specific cost recovery proce- families in the program who do pay for services. Govern- dures (which involve negotiations with local govern- ment reluctance to make defaulters pay may be a form of ments) have not been worked out, the participants have disguised subsidy to participants. Such a tactic may not been properly informed of their obligations, and so permit lower-income families to live in housing projects on. Once the cost recovery objective is cast in an they might otherwise be unable to afford, but that has appropriate management framework, it will be neces- not been the inevitable result. Subsidized programs typi- sary to use available management information to moni- cally have severely limited supply capacities, and in tor progress toward this objective, as with all others. situations in which output is artifically restricted, high- The El Salvador program offers a strong counterex- er-income families have proved adept in capturing the ample. Because of its small size and its nonprofit status, rationed benefits. In any case, the costs will eventually which requires it to achieve cost recovery to remain in have to be met, either directly or indirectly, by others. operation, the Fundaci6n Salvadorena de Desarrollo y Because the causes of default are complex-involving Vivienda Minima (FSDVM) has achieved an excellent re- affordability concerns, the adequacy of administrative payment record for its shelter program. Through mate- machinery, and the roles of political pressure groups-it rial help and collaboration on important community is difficult to isolate the contributions of individual fac- matters families are brought to understand that repay- tors to cost recovery problems. ments are necessary to the survival of project facilities. Serious cost recovery problems have arisen in the The FSDVM uses effectively a phalanx of incentives and Lusaka program: more than 50 percent of families in penalties to achieve cost recovery. These have included some upgrading projects are in arrears, and some house- support to community organizations in return for their holds do not appear to have paid anything. In this in- assistance in the collection of payments, careful screen- stance it has been possible to clarify causes. The initial ing procedures to select participants able to pay project Improving the Effectiveness of Urban Projects 173 fees, and visits by lawyers to families who have fallen The evaluated projects have witnessed both the posi- more than three months behind in their payments to tive and negative effects of active community involve- press them to pay. Eviction, although seldom used to ment and noninvolvement in project implementation. date, is an option which the Fundaci6n will not hesitate Because of its commitment to stimulating and using to use if necessary. community participation in achieving social goals, the These experiences reveal the usefulness of designing FSDVM in El Salvador has compiled a generally notewor- projects in a local public finance context, where the thy record of fostering feelings of responsibility for ability to operate and maintain infrastructure and ser- project facilities and services and for collections. Re- vices can be directly related to the resulting benefits as quired mutual help construction has also engendered well as to collections or cost recovery. From the outset of collective action toward social aims. Yet there have been a project, collection mechanisms ought to be designed several cases in which groups have organized against the with community participation in mind, and with regu- FSDVM and refused to make payments until basic services lar, relatively convenient payment schedules that are such as water were provided.'8 Community organiza- thoroughly explained. Participants must be made aware tions, it seems, cannot be expected to be simply passive of the direct relation between installment payments and or neutral variables in project implementation, for they loan repayments on the one hand and goods and services can act as stimuli or as deterrents to reaching objectives, received on the other, as well as what the consequences depending on the concordance of popular and program of default will be, both for themselves and for the project goals. as a whole. Methods for improving cost recovery per- The Lusaka project, too, has used community involve- formance which have been developed in other Bank- ment to its advantage in some cases and has failed to financed projects have included discounts for early re- employ it sufficiently in others. Through consultation payment and the withholding of materials or house with groups in upgrading areas and those scheduled to consolidation loans from families that default on lot be moved to overspill areas, major disturbances were installments. Collection methods based on property avoided. This accomplishment appears to have been an values have been used and could be extended. In sum, important one, in that squatter areas had traditionally whereas affordable projects and workable cost recovery been centers of unrest. Community groups, for example, methods can be designed without undue difficulty, the were organized at the planning stage into road planning real problems are in implementation. The require- committees, which made recommendations for mod- ments-in addition to political will-are better- ifications in road layouts (and the resulting dislocations integrated planning and considerably more attention of certain families) that were acceptable to all parties. during implementation. The committees have continued to be instrumental in other decisionmaking functions that affect community Community Participation well-being. Other components, including mutual help, have been less successful. Although community re- Finally, community participation in project imple- sources were successfully applied to the provision of mentation, and particularly in design, can have decided clinics, other efforts, such as one to provide multipur- effects on efficient achievement of objectives, including pose community centers, have not engendered popular maintenance and cost recovery objectives. Although support. popular support and consultation are indispensable for The Philippines evaluation team has noted that a eliminating resistance to project interventions, lengthy marked passivity characterizes group meetings where discussion of procedures or training of project promo- project plans are discussed. Tondo dwellers seem regu- tores in organizational skills, for example, can delay larly to follow barangay (community) leadership in de- implementation schedules substantially. In some in- claring preferences for particular services or procedures, stances project acceptability can be improved and costs even when these involve significant physical disruptions reduced if community groups accept direct responsibil- in reblocking their homes. Although this form of hierar- ity for the operation and maintenance of facilities, but chical organization, which involves consultation with devising new administrative structures and procedures barangay leadership and community groups, appears to for such purposes can be costly. Effective community have led, as in Zambia, to the circumvention of major participation is itself a goal of certain projects-for ex- social disruptions during the reblocking process, the ample those of the FSDVM in El Salvador-and hence evaluation team has recommended that more compre- some project managers may be willing to invest more hensive information campaigns be conducted before and resources in developing community institutions than during relocation so that households understand as are others who are more committed merely to attaining completely as possible what options are open to them. physical goals. The options may include actions which do not necessar- 174 Douglas H. Keare ily fit the preferences of barangay leaders, but inclusion In the realm of policy, project experience and evalua- of such choices may lead to a broader-based community tion findings confirm that housing markets are quite participation than the estimable grass-roots movement easily unfettered, and that once they are unfettered, they already achieved. respond vigorously. This argues for a strategy that makes it possible for more families to participate and that does not seek to constrain the development they Some Implications for Project Design choose to undertake. The first consideration implies and Policy Formulation making tenure, basic services, and housebuilding re- sources more easily available to lower-income families; The analysis of the eight project effectiveness compo- affordability and cost recovery findings indicate that this nents presented in the preceding section does not begin is a risk that can be run. The second consideration to exhaust the ways in which shelter programs might be implies the removal of restrictions (such as those on evaluated. Strictly comparable data have not been gener- rentals) that act, directly or indirectly, to suppress con- ated in every instance, given the wide variance in proce- struction. dures and in supply and demand features in the four The range of preferences implied by the above exam- country projects. This framework for gauging project ples argues unambiguously for housing solutions that efficiency nevertheless constitutes a useful tool for be- permit substantial flexibility to households. The optimal ginning to assess crossnationally the options and con- degree of flexibility is not known and may be undeter- straints in housing programs which attempt to serve the minable, but evaluation findings suggest that project needs of low-income and lower-middle-income urban planners should leave more to the market and resist populations. In this section some of the lessons for temptations to build too many components into proj- designing projects and for addressing wider policy issues ects. Rather, they should try to be selective and should are discussed. think of projects as series of interventions that may eventually prompt the system to produce all the desir- MGarket Forces and Flexibility in Design able sets of complementary components and services.'9 In this view, then, projects should be kept simple- A general feature of the findings is that market forces designed for realistic results and speedy imple- among the poor who seek shelter in rapidly urbanizing mentation.2" Key elements such as appropriate loca- developing countries are stronger and more easily re- tions, reasonably secure tenure, and efficient credit leased from constraints than had been anticipated. With- mechanisms must be in place, but beyond these the set in the limits determined by household economies, fami- of needed components may vary greatly. Far from de- lies exhibit varying behavior in housing markets (as is nying the complexity of urban problems, this approach generally corroborated by evidence from other projects). requires careful and complex analysis to relate objectives The planning and execution of the progressive develop- to resources and constraints, which leads to more accu- ment process, and the size, quality, and use of project rate choices about what to do first in a given set of houses, will vary enormously as well. circumstances. Along with continually improving provi- In the realm of project design, it has been shown that sions for maintenance, cost recovery, and community substantial opportunity costs render it uneconomic for participation, this approach may make possible more large proportions of target populations to utilize family rapid and more easily replicable improvements in hous- labor in constructing or improving their houses. It fol- ing for more people. lows that families should not be forced to use their own labor. Nevertheless, the initial decision to encourage use Affordability and Poject Design of own labor in self-help processes was a wise one. This conclusion is confirmed by the quality of houses con- Although affordability was not an explicit objective of structed using family labor in whole or in part and by the early projects, evaluation experience has shown that efficiency of the process for families with relatively low housing can be produced that is affordable for house- opportunity costs and relatively high levels of construc- holds in the lowest quintile of most urban income dis- tion skills. Similarly, the findings with respect to mate- tributions. If social equity objectives as well as efficiency rials stores, types of materials used, and the amounts of goals are to be met as directly and as rapidly as is rental accommodations constructed indicate that fami- desirable, efforts over 1985-95 will have to be more lies make a wide variety of choices, again usually for concerted than those already successfully carried out. good reasons. The proper interpretation of these results Though each particular set of circumstances dictates is that a broad range of solutions should be encouraged, ways in which performance may be improved, evaluation or at least allowed. findings suggest four principal ways of improving deliv- Improving the Effectiveness of Urban Projects 175 ery of project benefits to the poorest urban dwellers: months after reblocking. To a large degree this was revise standards, and thus entry costs, to lower levels; accomplished by increasing the amount of vertical increase the opportunities for rental arrangements in building: a quarter of the households (and half of those projects; tailor credit arrangements to the revealed that previously occupied single-story dwellings) added a needs of a population which may have substantial access second story for the first time, and the average number to informal transfers; and insist on excellent cost recov- of floors increased from 1.49 to 1.66. Projections on the ery performance in projects so designed. Urban project basis of early trends suggest that this figure may by now designers in the World Bank have been making consid- have reached about 1.85. erable progress along these lines in recent years. The most up-to-date information on a large urban The first two approaches to broadening the benefits of housing market is for Cairo,23 where the character of the shelter projects are not necessarily complementary and housing market has been changing rapidly as a result of need to be handled with sensitivity. Concerning stan- increasing demand that impinges on an inelastic supply dards, there are two crucial issues: the land density of of land in a large metropolitan area. A substantial share settlement, and service levels.2" The first issue is closely of the increase in housing has consistently been tied to plot size, although the two concepts are not accounted for by the informal sector.24 The expansion synonymous. Given the scarcity of urban land, particu- has occurred both by extension of the urbanized area larly in the very large metropolitan areas, urban projects and by the addition of stories to existing buildings. As a should have densities that are as high as possible, consis- result, informal housing that is initially constructed tent with the economic demands of the target popula- with one or two stories eventually reaches two to five tions and with reasonable health standards. Other things stories, and the average number of floors per building being equal, this strategy suggests small plot sizes- has increased from 2.09 in 1976 to 2.45 in 1981. The smaller, in general, than those considered during 1971- economics of this process are reasonably clear. The cost 75, when the projects evaluated in this program were of a medium-size informal lot in a medium-price loca- being devised. This does not imply less living space per tion in Cairo is now approximately two and a half to individual or per household, however. Plot size and three times the cost of constructing a 50-square-meter living area will vary greatly with the amount of vertical dwelling. Thus, conservatively speaking, unit costs per construction, which in turn will be determined by a square meter can be reduced by as much as a third by number of considerations, including land price, rental adding one story and by nearly half by adding two. market conditions, the availability of credit, and the Furthermore, the direct costs of an additional story will incremental costs of constructing additional stories. The be no more than 75 percent of the costs of buildinga new relevant factors, and thus the feasibility, profitability, single-story house on the minimum-size lot at the and amount of vertical building, vary from one locale to periphery.' another. Evaluation results indicate that a key objective of The early projects evaluated under the present pro- project design should be to bring (or keep) the unit price gram did nothing to encourage multistory construction, and quantity of land within affordable limits for the primarily because two of the four projects were in Afri- target population so that the poor are not overextended can cities of modest size where populations and land by participation or the affluent unduly encouraged to prices were not yet at levels that would force considera- participate.26 Fortunately, there are signs that signifi- tion of vertical building. In addition, multistory designs cant improvements in practices of the recent past are were explicitly avoided to minimize risks in the early possible. That there is a sound basis for such a favorable pilot projects. There is now considerable evidence, outlook is confirmed by project experience in El Salva- however, that builders in the informal sector are capable dor, where both the programs of the FSDVM and the use of of raising structures to substantial heights and will do so evaluation are of long standing. if appropriate incentives are present. The task of future Advances have been essentially of two kinds: improve- projects will be to see that the incentives are there and ments in layout, which have increased the propor- particularly that potential biases against multistory con- tion of residential area to total area, and stimulation of struction are avoided.' two-story construction. Improvements in layouts were Analyses of available data on the first upgraded areas achieved principally by significantly reducing the pro- to be reblocked within the Tondo area in Manila show portion of land assigned to vehicular traffic and parking that adjustments in vertical structure took place in short (through restriction of these functions to peripheral order. The average lot size was reduced by 12 percent areas) and by grouping houses around miniparks or (from 65.3 to 57.6 square meters), but households in- green areas which provide access to the interior plots creased the average constructed area by 34 percent and serve as semiprivate recreation areas. Families thus (from 40.2 to 54 square meters) within the first three trade off some private area for shared use of areas. The 176 Douglas H. Keare success of these design choices is illustrated by the fact sponding financial obligations. Each possibility must be that residential area as a percentage of total area has treated squarely in project design. increased from an average of about 50 percent in the There are several related considerations regarding earliest projects in El Salvador to 70-80 percent in the service levels. Water supply provides a somewhat sim- most recent ones.27 plistic illustration. The questions in this case will be At the same time, there has been a steady increase in whether to have individual household connections or El Salvador in the number of units per hectare of res- some kind of communal supply (such as standpipes) and idential area. This figure has slowly increased as plot when to introduce individual connections for a particu- sizes have been reduced. The greatest increase in unit lar populace.29 The problem is analogous to that of core densities has come in an experimental project (Cona- housing: whenever possible, options should be offered. caste) in San Salvador, where a two-story design for Because not all households can afford individual connec- some of the units has permitted the construction of 140 tions, access to some form of communal supply should units per hectare, as against an average of 80-95 units in be an option. What proportion of participants requires other recent projects. The cost-effectiveness of this re- this option and what the range of options should be are sult appears quite compelling for large urban centers essentially empirical demand questions that are not eas- where land prices are sufficiently high to encourage the ily assessed.3" The planning question is one of designing substitution of capital (construction) for land. This ex- economical total reticulation systems (including fire ample implies that a two-story unit (25 square meters of hydrant and waste disposal systems) over time. constructed area on a 32-square-meter lot) costs about 15 percent less than a 25-square-meter single-story The Rental Option and Other Considerations house on a 60-square-meter lot. Furthermore, rising land prices are inexorably shifting the balance further in Even with the provision of options and careful control favor of multistory construction and making such con- of costs, a substantial portion of families, particularly struction economical over more of the urban landscape. among the poorest, will have access to project shelter Of course, Salvadorean cities are not yet large enough to only to the extent that rental accommodations are ex- provoke the strong incentives for vertical construction panded along with ownership opportunities. This con- evident in the examples for very large cities such as Cairo clusion is based on the fact that a significant proportion and Manila. of households in all of the countries covered by the Two other design refinements which can be used to evaluation rents rather than owns housing. This finding reduce unit costs are the elimination of unnecessary does not imply that all those renting do so as a matter of construction costs and the omission of undesired ser- preference. The choices in a given situation depend on a vices or service levels. These considerations too are best particular confluence of supply and demand and will be approached from the perspective of the diversity of de- biased toward rentals in the circumstances of restricted mand and require the preservation of a range of options supply and overcrowding that prevail in many, if not as well as flexibility in their use. The design of core units most, developing-country cities. The point is not that can be used to illustrate the choices involved. General rental accommodations and ownership opportunities judgments concerning the desirable extent of the core should be expanded in step with each other, nor even will depend on the relative benefits and costs of the that project planners should concern themselves with possible construction methods.28 Precise estimation of precise ratios, but only that in most instances both these costs and benefits is an aim of current research, should expand.3' The evaluation research has not indi- and the findings should be factored into small unit cated any particular need to stimulate the rental market, models, such as the Bertaud model cited in note 27, to although appropriate improvements to credit systems yield the requisite planning guidance. A further point to (embracing loans to contractors as well as to purchas- be emphasized is that what is shown to be desirable on ers) would increase the elasticity of supply of rental average is not necessarily desirable for all concerned. accommodations. The most important action is to en- For example, even if calculations showed that core hous- sure against avoidable constraints on rental arrange- ing has on balance an unfavorable benefit-cost ratio, a ments, such as the restrictions and prohibitions de- significant proportion of potential residents might ben- signed into some of the early projects financed by the efit from the option of core housing. Conversely, World Bank. although the overall benefit-cost ratio might be favor- There is apparently some degree of resistance on the able, that result should not obscure the fact that a part of developing-country policymakers to rental op- significant number of potential participants would ben- tions, and project planners must take that resistance efit from options other than core housing and its corre- into account. The underlying notion that renting does Improving the Effectiveness of Urban Projects 177 not contribute to suitable solutions to urban shelter the scope for improving the total housing stock progres- problems seems untenable in light of the evidence that sively over time, rather than on the fine-tuning involved large proportions of households in all countries choose in selecting particular types of projects. to rent. The notion expresses a fear that the renting poor Along with these changes in design, which lead to will be exploited by a rentier class-a phenomenon that simpler and lower standards and costs and an appropri- exists to some extent everywhere. The best way to mini- ate balance between ownership and rental options, mize this tendency is to expand the total housing supply should come concomitant-changes in credit facilities. and ensure, by monitoring for specific abuses, that full Not much is known yet about the real demand for credit, cost recovery is enforced so that projects do not subsi- and experimentation has only begun, but evaluations of dize a rentier class.32 A well-executed strategy of this experience do permit the statement of a few principles. kind will avoid the feared outcome much more success- Not all participating families will require institutional fully than such misguided policies as rent controls. credit, and the design of credit programs should be Finally, it will be necessary to experiment with means directed to those that do. The size of credit programs (other than renting out housing space) of enabling the should be reckoned in terms of the numbers who need poorest households, particularly, to utilize houses and credit and their requirements rather than a rationing of plots as earning assets, for commercial or small manu- supplies among the entire population. Credit should also facturing enterprises or for growing food. be apportioned to builders who contemplate rental ac- To reiterate, flexibility is crucial in seeking the mix of commodations, not only to owner-occupiers. components that will serve the housing needs of target urban populations. The outlook for progress along these lines is encouraging. For example, at the same time that Cost Recovery it adopted the sites and services concept, the World These remarks lead to a final, related observation: Bank, realizing the limitations of that concept, began to shelter program maintenance, collections, and cost re- develop the complementary area upgrading approach. covery-and their interrelations-should be vital con- Recent years have also brought a significant broadening siderations for designers and managers. They should be of options and modification of concepts under both dealt with from the beginning of project identification, approaches, leading, for example, to a change from and projects should not be executed that do not have a materials loans to construction loans, to acceptance of favorable relation between revenue extracted and ben- the need for rental components, and to numerous other efits delivered as well as between benefits and costs. For adaptations. The best illustration of such adaptability replicability, programs must be designed so that they thus far has been in El Salvador, where the FSDVM has return, directly and indirectly, approximately as much modified its mutual help program significantly, pushed to public revenues as they require in expenditures." its already low standards still lower in some projects, and Beyond this, more must be done to make policymakers developed the layout and design innovations mentioned aware of the potential that projects offer for improving above. Analogous changes have been made in the de- the local fiscal situation. In particular, projects should veloping programs of the National Housing Authority be used to prompt fuller cost recovery not only among (NHA) in the Philippines.33 the target groups themselves but, even more important, Analyses aimed at assessing the relative merits of the among higher-income strata if, as in Zambia and many area upgrading and sites and services approaches have other countries, their housing and services are being suggested that, far from being distinct strategies, they subsidized. belong to a continuum of possibilities with countless It is not claimed here that this is an easy task. On the nuances. The processes involved in restructuring a contrary, precisely because it is difficult, further ex- wooden house on a Tondo lot, repositioning it within a perimentation is necessary, and planning should include new grid established by reblocking decisions, moving it considerable experimentation with various forms of to an adjacent overspill area, or even removing it to the community participation and contribution. Fortu- Dagat-Dagatan sites and services area four kilometers nately, the record on this issue is encouraging. Just as away are at least conceptually similar. To be sure, time, they have a desire and capacity to improve their own distance, cost, and the extent of retention of the original homes, the urban poor have a desire and capacity for structure all vary. The point to be stressed is that they greater participation in their own governance. In the vary along a continuous frame of reference or response past these capacities have been left largely untapped by and that the differences are frequently very small. Leav- program designers, but it has been demonstrated that ing aside the extremes, the principal effort should be to projects can make effective use of them. More can be focus policymakers' attention on this continuum and on done, however, to facilitate project implementation, im- 178 Douglas H. Keare prove the record on collections, and articulate demands process evaluation techniques. More recently, the World for specific components in future projects within an Bank initiated a first experiment that employs partici- overall strategy for progressive social change and de- pant observers in the field. velopment. Judicious involvement of the community at An illustration of improved and extended measure- all these levels must be more fully explored in planning, ments is the demand studies carried out in El Salvador project design, evaluation, and associated research. and the Philippines under the Evaluation Program. Along with similar studies in Korea and Colombia under the World Bank's Research Program, these have helped The Contributions of Evaluation to lay the groundwork for a subsequent generation of urban projects which, by virtue of improved estimates, All of the above findings have stemmed in large mea- may be designed and priced more nearly in accordance sure from the World Bank's Evaluation Program.' Eval- with the demands of the intended beneficiaries. Also as a uation and the associated research have contributed to result of this work, additional research is under way in the advancement of project design and policy formula- these and other countries to extend the range of cir- tion in four principal ways. First, evaluation has cumstances (income, climate, city size, culture, and so strengthened the modeling paradigms which frame the on) for which there are reasonable demand estimates. In design of both projects and research. Second, it has addition, the pioneering work in this program on hedon- contributed over time to better identification of the uses ics analysis as applied to housing has demonstrated the and users of all management information, including overall cost-effectiveness of the progressive develop- that provided by evaluation. Third, it has contributed to ment approach at the same time that it has begun to the refinement of indicators and measurements. Fourth, develop information on the relative contributions of it has helped to break new ground with respect to certain various components to overall project benefits. relations between key variables. Each of these contribu- The Evaluation Program can truly be said to have tions will be discussed briefly. pioneered the investigation of transfers within the ex- By providing a rigorous framework, evaluation has tended family and other kin and nonkin networks. That contributed to keeping a framework in place notwith- such transfers can be important was first discovered standing alterations that were desired or required as somewhat incidentally in a study of control groups in project and program lessons were learned. By contribut- Dakar, Senegal, in 1977. The phenomenon was thereaf- ing both to the alterations and to the stability of the basic ter followed closely in all projects, particularly in the edifice, evaluation has helped to bring the framework Tondo project in Manila, the Philippines, where the into clearer focus. Examples of such contributions to budgets of a small sample of families were followed better understanding include putting self-help and closely over three years. Data from the evaluations in El mutual help into appropriate perspectives, identifying Salvador, the Philippines, and Senegal, as well as from a the roles of rentals and transfers in the development related investigation in Cartagena, Colombia, have been process, and spelling out the needs for designing better used in several research papers. These have demon- housing finance milieux and credit schemes. strated, among other things, that transfers are very In the realm of identifying uses and users, evaluation important in the budgets of the poor, making up close to has operated to narrow what was a sizable gap at the half the total incomes of half the families in the lowest outset. The initial evaluation design, which was based on third of the income distribution; that they tend to flow quasi-experimental design and statistical rigor and from better-off to poorer households within the spec- which sought to address a large proportion of the rel- trum of the urban poor; and that they tend to be sensi- evant questions, was extremely ambitious and excessive- tive to household composition and employment status ly demanding. It was in everyone's interest to seize (households headed by females or unemployed workers opportunities to improve the focus and become more are more likely to receive transfers than those headed by selective as the program progressed and each cycle of employed males) as well as to shifts in income. Further- findings was reviewed. Equally important, process eval- more, it appears that there is a basic underlying contract uation, which did not play a major role in the beginning, by which resources are provided to poorer households to was blended in progressively over time. Panel studies, meet basic needs or to improve housing when opportu- which did form part of the original design, were used nities such as these projects afford are present. Current- more as time went on. As evaluation results attracted ly the research is being extended outside the evaluation managers' attention, the managers asked evaluators to framework to exploit the best data bases and the most study and report on increasing numbers of immediate encompassing sample frames available. This work also problems. This required increased use of and experience addresses the important issue of urban-rural transfers. with evaluability assessment, rapid feedback, and other It is worth noting in closing that these concluding Improving the Effectiveness of Urban Projects 179 statements can be made and documented in large mea- El Salvador projects evaluated, weekends were the least favor- sure because not only the projects being evaluated but able times to participate in mutual help, as this was when their also the evaluation program itself have been under con- (informal sector) earnings were typically the highest. stant scrutiny. Thus, evaluation methods and processes 10. Use of the term demnand has been avoided up to this were evaluated along with evaluation results at each of point, forvery little is knownabout it. Thiswastruewhenthese six annual conferences in which project managers, field projects were designed, and although the research program sixvannualors, confrespncesinl wfich s projt m onaersfied has contributed materially to great advances in the under- evaluators, and responsible officials of the sponsoring standing of the demand for housing in recent years, the (de- agencies participated. A detailed record of these dis- rived) demand for credit has been little studied to date. cussions and the consequent decisions has been kept. 11. Furthermore, the evaluation's findings about income This made possible the meta-evaluation, or "evaluation transfers indicate that the two types of restriction may be of evaluation programs," on which these concluding incompatible. That is, since families which are alike in other remarks have been based. respects have very different access to transfers, they must be supposed to have commensurately different demands for credit. If this is so, a given loan fund, whether or not adequate in itself, will not be optimally allocated among families by rationing a small amount to each. Notes 12. The evaluation findings are so far consistent with this finding (see Jimenez 1982). In addition, the administrative 1. Progressive development is a method of housing con- mechanism for ensuring materials purchase may carry its own struction or upgrading that is achieved through staged de- restrictions and added costs. velopment over extended periods and involves considerable 13. This paragraph encapsulates the frustrations experi- flexibility in housing design, materials used, and the family's enced in the Senegal project, the only one of the four to (self-help) contribution to the construction process. experience serious difficulties. 2. At the time, experience with such approaches to shelter 14. A social problem, however, may still exist. If significant developmentwas sparse, and the bulk of information-most of numbers of families that have moved to new sites fail to build it, at this stage, speculation-came from observation of spon- permanent structures within a reasonable period, it may be taneous "invasions" of land by squatters in several countries. difficult to obtain and sustain policymakers' support for the Empirical research had been extremely scanty, and rigorous sites and services option. evaluation of experience was nonexistent. 15. Another way of viewing this tradeoff is that if a family is 3. It has not been possible as yet to analyze whether present forced into this situation and chooses nevertheless to con- levels of project costs and standards systematically contribute struct a house with its own labor, the cost of this self-help to underrepresentation of lower-income households from the method should include the cost of rent during the construc- start of projects. tion period-which, in El Salvador, has run up to six months. 4. In some interior cities of El Salvador lower population This added cost would not be present, or present to the same densities and good access to well water have made the larger extent, in some of the "higher-cost" options, for example, lots plots and fewer services of the colonias more attractive than with core housing. the options designed by the project agency on the basis of its 16. In this connection it is noteworthy that in the same earlier experience in the larger cities. project consideration has been given to vesting the responsibil- 5. Reblocking is the process of applying, insofar as possible, ity for maintenance of communal standpipes, where similar normal subdivision processes to the less orderly division of problems could be anticipated, in community groups. It has space that results from most illegal occupations of land by also been suggested that if a separate garbage collection fee squatters. The physical subdivision process requires a counter- existed, garbage collection services might be improved. The part process of mapping prior "claims" in the old order into adequacy of service would then affect, and be affected by, accepted titles in the new. payment of this fee alone and not the entire cost recovery 6. Explicit selection does not occur within area upgrading effort. projects, since they typically deal with pre-existing popula- 17. For additional details on project collection efforts and tions. on the constraints on the LCC's handling of arrearages, see 7. In El Salvador and Zambia, where the genuine concern of Bamberger, Sanyal, and Valverde (1982), chs. 9 and 10. project administrations that the lower income groups be 18. In one case the FSDVM promoted, or at least actively reached predated any involvement by international agencies encouraged, a protest directed against the recalcitrant state such as the World Bank, the selection procedures appear to water utility. This demonstration inadvertently provided a have had a particularly good record of selecting families from model for organizing pressure on the Fundaci6n itself later. within prescribed income ranges. 19. A model to guide practice in the adaptation of project 8. This discrimination has been corrected insofar as components to actual demand is not yet available. A high possible. priority should be given to developing one. 9. What is more, contrary to what one might expect from 20. As pointed out above, there were substantial delays in experience in developed countries, where formal sector em- the implementation of projects already evaluated. More recent ployment predominates, for the majority of participants in the experience indicates, however, that implementation delays for 180 Douglas H. Keare urban projects are very near the mean for all Bank-financed indirect effects on housing markets of changes in supply projects. wrought by large projects need to be performed within evalua- 21. The timing of the introduction of various services is a tion research. third, ancillary consideration. 33. Few specified examples can at present be cited for the 22. Policymakers are sometimes reluctant to promote or other two countries in the evaluation program. This is princi- even to permit two-story construction on the grounds that a pally a result of their later phasing and slower development and significant proportion of families would be unable to build the absence to date of second projects. Nevertheless, a similar "suitable" additional stories. Evaluation findings that show the pragmatism has been evinced in response to problems encoun- extent to which families hire others to do their building should tered in the implementation of the first projects. help to redefine this risk. 34. This does not imply that there should be no subsidy 23. Mayo (1982). Comparable information is becoming from higher-income groups, either within projects or over a available under a World Bank research project on one or more broader spectrum, but rather that the subsidy should be on a cities in Colombia, El Salvador, India, Indonesia, Korea, and scale that can be sustained in a large program over time. the Philippines. 35. This is not to claim that operational personnel would 24. Informal housing-defined for the purposes of this have remained in ignorance without a rigorous evaluation study as housing which has been constructed without official program. Sometimes the first insights have come from the permission to convert land from agricultural use, without a evaluation program, but just as often evaluation has merely building permit, or in violation of building codes-constituted confirmed suspicions that arose from project experience. The about 90 percent of new housing starts in 1971-76 and 75 key point is that, by virtue of the rigor of its approaches, percent after 1976. evaluation has assisted the learning process to proceed more 25. Infrastructure costs for building to these implied densi- confidently-and that is an important contribution. ties may, however, reduce the advantages of the central loca- tion. Bbigah 26. Although the plot sizes of past projects have generally Bibliography been acceptable to the target populations, little information is available as to optimal plot size distributions in projects. This Bamberger, Michael, Edgardo Gonzalez-Polio, and Umnuay important area should be given priority in evaluation and Sae-Hau. 1982.Evaluation ofthe First El Salvador Sites and related research, particularly in the developing world's largest Services Project. World Bank Staff Working Paper 549. and most rapidly growing cities. Washington, D.C. 27. Such success has been one factor that has motivated the Bamberger, Michael, Bishwapura Sanyal, and Nelson Valverde. World Bank to develop small unit planning models to deal 1982.Evaluation ofSites andServices Projects: The Experi- efficiently with tradeoffs in project layout design. See, for encefromLusaka, Zambia. World Bank Staff Working Paper example, PADco and the World Bank (1981). 548. Washington, D.C. 28. Such calculations will include the benefits, if they can Jimenez, Emmanuel. 1982. "The Economics of Self-Help be demonstrated, of providing some amount of core housing to Housing: Theory and Some Evidence from a Developing minimize reticulation costs, encourage use of sanitation facili- Country." Journal of Urban Economics, vol. 11, no. 2 ties, or stimulate construction by making it easier to move (March), pp. 205-28. in at once. Kaufmann, Daniel. 1982. "Social Interactions as a Strategy of 29. There are, of course, many other factors which affect Economic Survival among the Urban Poor." Ph.D. disserta- unit costs, such as capacity, hours of service, and quality of tion. Harvard University, Cambridge, Mass. Processed. water. Attention is limited here to factors which directly affect Keare, Douglas H., and Emmanuel iinez. 1983. Progressive project area design.Kae oga . n maulSmnz 93 rvesv 30.oDemandcanalysestfrom all four countries in the original Development andAffordability in the Design of Urban Shel- 30. emad aalyss fom ll our ounrie in he rignal ter Projects. World Bank Staff Working Paper 560. Washing- evaluation program show that significant proportions of the ter P .Wa target group are satisfied with communal water supply and pit ton, D.C. latrines (Zambia project participants), choose low service level Keare, Douglas H., and Scott Parris. 1982. Evaluation of Shel- options if they are available (El Salvador and Senegal control ter Programs for the Urban Poor: Principal Findings. World groups), and place a relatively low hedonic evaluation on Bank Staff Working Paper 547. Washington, D.C. sanitation facilities (Philippines project participants). Mayo, Stephen. 1982. "Informal Housing in Egypt." Cam- 31. There may be instances where crowding has been of bridge, Mass.: ABT Associates. such proportions that a significant expansion of owner- PADCo and the World Bank. 1981. The BertaudModel: A Model occupied housing will imply that the absolute number of for the Analysis of Altematives for Low-Income Shelter in renting families decreases. the Developing World. World Bank, Urban Development 32. To guide this aspect of the program, analyses of the Department Technical Paper 2. Washington, D.C. 15 Urban Economic Policy: Directions for the Future George S. Tolley and Vinod Thomas T he past quarter-century has witnessed rapid urban- forces of urbanization that originate from population ization in the developing world. Today the develop- growth, urban-rural productivity, and composition of ing countries contain fifteen of the twenty-five cities demand increases are also widely applicable. The book with populations over 5 million. Although urbanization has laid out some of these factors in the developing- is associated with economic growth in the minds of country context to assist in an understanding of ur- many, the scale of urbanization in developing countries banization trends. The authors differ in their emphasis is also viewed with alarm. The observed concentration of on the relative importance of these forces, perhaps be- poverty, unemployment, housing, and transport prob- cause the experiences of the countries examined vary. lems in some highly built-up urban centers such as Some of the differences between developed and de- Mexico City, Sao Paulo, Bombay, Cairo, Jakarta, and veloping countries are perhaps more interesting than Manila has given rise to fear of the consequences of the similarities. Even if the tools of analysis are not present trends. necessarily different, the present-day urbanization ex- The rate and scale of the recent urbanization phe- perience in the developing world brings to focus more nomenon have been so large that they are not yet ade- complex and difficult, if not different, policy issues. quately understood. Urban theory offers explanations of Urban problems in developing countries are more severe the workings of urbanization in developed economies: than those in the developed countries today because of theories of housing markets, locational decisions, traffic the former's much faster rate of urbanization at present. and transport modes, and labor markets have been ad- In addition, the problems are also bigger than those that vanced. Are the same analytical approaches applicable in the advanced economies experienced when they were developing countries as well? Is developing-country ur- developing. The main reason is not that the pace of banization a distinct phenomenon? Are there urban pol- urbanization is different now than earlier, but that ur- icy conclusions for developing countries that differ ban populations and agglomerations are larger today, markedly from those for the industrial economies of partly because of larger overall population. Moreover, in today? developing countries today intraurban transport costs This book has emphasized similarities in urbanization are relatively lower than those in previous time periods in developing and developed countries but has also and allow flatter rent gradients within cities, larger city brought out important differences. Many of the analyti- sizes and more concentration. (Reductions in intracity cal methods adopted in this work were developed in the transport costs may have been greater than those in context of experience in advanced countries. Basic be- intercity costs.) Urban centers in developing countries havioral and economic explanations behind rent and also seem to have been able to achieve greater technol- density gradients and city formation are relevant in ogy transfer and productivity gains than rural areas, developing as well as developed countries, and even the which has caused correspondingly stronger pressures to range of parametrical estimates of some key demand urbanize in an increasingly interrelated world economy. variables in the urban economy may be similar. The Individual countries obviously face widely differing cir- 181 182 George S. Tolley and Vinod Thomas cumstances; factors range from low and stagnant agri- source of employment and incomes which should not be cultural productivity in many African countries to spe- discouraged. cial city incentives and government urban subsidies that Antipoverty efforts in urban areas are best served over provide added impetus to urbanization in many parts of the long term by employment and labor market policies the world. that increase labor productivity and demand more rap- Although the urban economy is exceedingly complex, idly than labor supply, with special emphasis on lower this book has adopted the view that an understanding of skill levels. Policies to assist labor mobility across occu- econornic behavior in the urban context is more useful pations and locations are also vital. In this context the than is the construction of large-scale urban models. urban employment and poverty record can be improved Much more comparative analytic work remains to be in many developing-country settings by measures to done, however, in promoting a better understanding of reduce the natural growth of population and to improve developing-country urbanization. The urban-rural link- labor force participation and education. Massive rural- ages vary among countries, as do the importance of urban migration is often held responsible for urban macroeconomic policy on urbanization and the impact poverty and unemployment, but the migration issue is a of regional considerations that affect urbanization. complex one in which urban-urban outflows play as Finally, international trade is an important determinant important a role as rural-urban migration. The eco- of the nature and composition of production and popula- nomic incentives to migration derive from the economy- tion within a country. Understanding these phenomena wide context, and measures at an individual city level- in greater detail constitutes an agenda for further work. such as restrictions on informal activities and squatter settlements and withholding of urban infrastructure for new entrants-are likely to be ineffective and inefficient. Urbanization Problems Urban subsidies for production and consumption often in Mixed Economies serve to increase city inflows, and a reduction in such special incentives could help to keep migration flows at Urban problems are best understood in the context of manageable levels. the economic development of nations. Urbanization as such is neither the source nor the enemy of develop- ment. The mainsprings of productivity growth make Lessons under Central Planning agglomeration and urbanization attractive and bestow net economic benefits, at least for some time. Urbaniza- China's experience illustrates the difficulty of in- tion can proceed too rapidly, however, and make some fluencing income gaps between the city and the developing-country cities too big from the economic and countryside. China has not relied on migration and management points of view. Some important aspects of seems to have worked much harder on more direct the overurbanization issue are negative externalities, measures than other societies. Through the encourage- such as crowding, pollution, and transport failure, ment of bootstrap operations villagers have been in- which represent a drain on the benefits associated with duced to help build an infrastructure of waterworks, urbanization. Overurbanization, if it occurs, is difficult roads, level fields, schools, public health care, and ad- to correct and may best be addressed through actions ministration. The government has tried to help directly aimed directly at the externalities rather than at city by raising prices for agricultural products, but a closer size. examination shows that the drive toward rapid industrial Often urban poverty is identified as the urban prob- growth has helped shape government investments and lem, and quick-fix solutions are sought to eradicate it. industrial prices to the detriment of agricultural growth. The basic reasons for urban poverty, however, relate to And when it comes to state-subsidized services, includ- overall poverty in the country and to inadequate gainful ing housing, health care, education, and cheap food, it is employment. Nationwide policies for generating long- still the urban sector that has reaped the greatest share term income and employment and for promoting a of the benefits. That these conditions are found in many smoother functioning of labor markets are perhaps the other developing societies seems to illustrate the con- best strategy for reducing urban poverty. Sometimes tinuing problems of urban and industrial bias even in a informal sector activities, such as shoestring enterprises society that has overtly renounced such a policy. and street vending, are falsely viewed as unproductive, China has now begun to correct some of these difficul- and their elimination is associated with poverty allevia- ties. The government has helped induce a rapid rise in tion. Again, the level of economic development of the rural incomes over the past few years by again raising country determines the mix of employment opportuni- rural purchase prices, by increasing agricultural invest- ties, and informal activities could represent a valuable ment, and by removing several other bureaucratic con- Urban Economic Policy: Directions for the Future 183 straints on farmer activities. The rise in rural incomes approach from an economic viewpoint. Experiences may still not be greater than the increase for urban with the concentration incentives in Brazil and the de- dwellers, but absolute living standards have begun to centralization inducements in Korea and Venezuela improve significantly. Some of the gap between city and which are documented in this book lend support to this countryside may be erased less by helping agriculture view. itself than by allowing peasants to open more small The indirect urban effects of economic policies are industry in rural market centers. This phenomenon has often ignored and are not well understood. Many coun- not been discussed in this book because many of the tries in Asia and Africa have historically adopted price market centers are excluded from China's urban statis- policies that hurt agricultural productivity and thereby tics. Yet with the prohibition on migration to cities, indirectly favored urban production. Consumer sub- these small market centers seem to be providing an sidies in urban areas have been widespread. Macroeco- outlet for peasants which exerts a positive influence on nomic policies have also indirectly and unintentionally rural incomes and possibly leads to a diffuse pattern of promoted a concentration of production of nontradable urbanization throughout the countryside. goods in urban areas. An overvalued exchange rate im- More peasants may also be slipping into cities. With plicitly taxes tradables and shifts resources to nontrad- the increasing role of peasant free markets in cities, the ables. To the extent that the agricultural sector contains reduction in the number of rationed goods in some relatively more tradables than does the rest of the econ- cities, new construction activities that can use unskilled omy (including services), on average, such an exchange labor, and a general relaxation of bureaucratic rule, the rate policy would promote nonagriculture in relation to number of illegal migrants in cities may be growing. agriculture. Increasing foreign trade may also create pressure for A reversal of the process of urbanization is unlikely to such major port cities as Shanghai to grow in away that be feasible, nor is it desirable as national policy if eco- they have not in the past. All these forces may lead to a nomic development is to be sustained. At the same time, slightly different Chinese urban structure than we have it should be noted that a country might wish to pursue seen in the past three decades, but much of the basic some type of decentralization of economic activities for structure is likely to remain intact. Centralized control reasons not readily measurable in economic terms. over investment, state jobs, and many rations continues Often, large concentrations of people and activities pre- to give the government a way of controlling the growth sent political problems. Another motive is that a larger of cities. And much of the new growth of rural market number of smaller cities may be more desirable, from a centers would promote a more balanced pattern of defense and strategic viewpoint, than a few densely urban growth, similar to that witnessed in earlier populated large cities. One problem is how to take decades. account of these goals legitimately rather than letting them be used as mere rationalizations for undesirable or ineffective policies. Urban Concentration and Urban Policy A relevant issue seems to be whether and how to marginally affect the degree and nature of urbanization, The distribution of city size is a subject of intense including the size distribution of cities. Clearly, politi- public debate in countries with divergent economic cal, social, and economic pressures influence policy op- managements. The issue-a variant on the concern tions, and policy choices are seldom straightforward. about urban concentration-relates mostly to the Nevertheless, the analysis of urbanization in this book alleged polarization of economic activity in one or a few yields four types of suggestions for economic policy. urban centers. Such concentration does often exist, as in First, policies might attempt to take account of exter- many Latin American countries; it is less marked in nalities that inadvertently affect urbanization. External- India. Considerable work has been done on the clas- ities-whether negative or positive-are not usually sification of cities according to the structure of the accounted for in the activity of the private sector. Public economic base, but a classification of desirable city size policy might correct this failure by, for example, adopt- distributions has yet to emerge. ing appropriate pricing policies for public services and Nationwide planning for a certain city size distribu- taxing public ills such as pollution and congestion. Such tion does not seem promising, even if such an outcome interventions are likely to influence the locational might look attractive. Even limited efforts to create choices of people and economic activities. cities and growth poles have turned out to be excessively Second, government policy could address vital infra- costly, as have policies to promote concentration. A structural issues. Bottlenecks in transport and other macroeconomic policy posture that does not favor or infrastructure constrain urban development across re- penalize urbanization seems to be the most desirable gions and affect the size distribution of cities. The key 184 George S. Tolley and Vinod Thomas issue concerns the optimal level and composition of a crucial question. In better understanding this issue, investment in infrastructure. In this connection, im- knowledge from other disciplines needs to be combined proved benefit-cost evaluation techniques and efficient with economic considerations, and intermediate goals application of policies would be essential. must be contrasted with those connected with long- Third, direct incentives to influence urbanization and term economic development. city size might be introduced only selectively in cases where net benefits can be shown. Policy experience to Bibliography date is not encouraging. The challenge, therefore, is to determine in which instances direct government in- Findlay, Sally. 1977. Planning for Internal Migration. tervention to promote or retard city size is warranted. A Washington, D.C.: U.S. Bureau of the Census. fundamental need is the development of rigorous and Ingram, Gregory K., and Alan Carroll. 1981. "The Spatial reliable means of making quantitative estimates of the Structure of Latin American Cities." Journal of Urban Eco- effects of these policies. nomics, vol. 9, no. 2 (March), pp. 257-73. Finally, several urban problems require direct solu- Linn, Johannes F. 1983. Cities in the Developing World: Poli- tions. Policies for greater efficiency in urban housing, cies for Their Equitable and Efficient Growth. New York: transport, finance, and other areas are often needed. The Oxford University Press. experience with large-scale housing projects and capital- Mazumdar, Dipak. 1976. "The Urban Informal Sector." World intensive transit systems has been unpromising. The Development, vol. 4, no. 9, pp. 655-79. evidence in this book supports smaller and more divisi- Mills, Edwin S., and Charles M. Becker. 1986. Studies in Indian ble projects which respond more directly to the demands Urban Development. New York: Oxford University Press. and financial means of consumers. Expenditures on Mohan, Rakesh. Labor Force Participation in a Developing rehabilitation, operation, and maintenance of urban Metropolis: Does Sex Matter? World Bank Staff Working facilities, in addition to capital deepening, can have Paper 749. Washington, D.C. particularly high payoffs. Efficient financial and eco- Renaud, Bertrand. 1981. National Urbanization Policy in De- nomic management of cities as urban centers reach veloping Countries. New York: Oxford University Press. extraordinary sizes is an increasing challenge. The rela- Schaefer, Kalman, and Cheywa Spindel. 1976. Sdo Paulo: tion between urban projects and policies on the one Urban Development and Employment. Geneva: Interna- hand and management efficiency in cities on the other is tional Labour Office. The most recent World Bank publications are described in the catalog New Publications, which is issued in the spring and fall of each year. The complete backlist of publications is shown in the annual Index of Publications, which contains an alphabetical title list and indexes of subjects, authors, and countries and regions; it is of value principally to libraries and institutional purchas- ers. The continuing research program is described in The World BankResearch Program:Abstracts of Current Studies, which is issued annually. The latest edition of each is available free of charge from Publications Sales Unit, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66, avenue d'lena, 75116 Paris, France. I THE WORLD BANK Urban problems, already acute in the developing world, threaten to become more severe as popula- tion growth and rural-to-urban migration continue. To promote a better understanding of urbani- zation, this volume offers a framework for analyzing urbanization processes and evaluating policies to cope with urban problems. Among the topics considered are the causes of rural-urban migration, the connection between urbanization and economic development, urban poverty, differences in country experiences, the goals and effects of policies that attempt to influence urbanization, and urban management issues in such areas as housing, transport, finance, and pollution. Approaches in both market and centrally planned economies are examined. Written by experts from the World Bank and the academic world, the studies build on progress to date and make new contributions that should be of value to policymakers, project analysts, scholars, and, in general, persons interested in urban problems. Introductory and concluding chapters by the editors draw together the findings and bring out policy implications. George S. Tolley is a professor of economics at the University of Chicago. Vinod Thomas is chief of the Country Analysis and Projections Division, Economic Analysis and Projections Department, The World Bank. WORLD BANK PUBLICATIONS OF RELATED INTEREST Metropolitan Management: The Asian Experience K. C. Sivaramakrishnan and Leslie Green Published for the World Bank by Oxford University Press Studies in Indian Urban Development Edwin S. Mills and Charles M. Becker Published for the World Bank by Oxford University Press Cities in Conflict: Studies in the Planning and Management of Asian Cities John P. Lea and John M. Courtney, editors Cities in the Developing World: Policies for Their Equitable and Efficient Growth Johannes F. Linn Published for the World Bank by Oxford University Press Urban Land Policies: Issues and Opportunities Harold B. Dunkerley, editor Published for the World Bank by Oxford University Press National Urbanization Policy in Developing Countries Bertrand M. Renaud Published for the World Bank by Oxford University Press ISBN 0-8213-0786-X